{ "schema": "raw-envelope-fixed-metrics-public-addendum-v1", "experiment_label": "2026-10-19", "role": "Append-only follow-up to the saved acoustic checkpoint; the earlier missing-value snapshot remains unchanged.", "state": "completed_with_missing_observations", "completed_rows": 90, "measurement_started_utc": "2026-10-18T17:45:53.038363+00:00", "measurement_completed_utc": "2026-10-18T17:55:30.475506+00:00", "scientific_units": { "recordings": 5, "speakers": 4, "conditions": 90, "conditions_per_variant": 30, "seeds_are_paired_repeats_not_independent_recordings": true, "repeat_executions_not_pooled": true, "inferential_statistics_performed": false }, "metric_units": { "row_relative_f0_mae_cents": "cent", "row_unvoiced_miss_rate": "fraction", "row_recognition": "CER rate difference from original, fraction; full CER and WER counts and rates are retained in detail", "comparison_automatic_mos": "predicted MOS difference to same-pitch same-seed PML", "comparison_relative_f0_mae_cents": "cent difference to same-pitch same-seed PML", "comparison_unvoiced_miss_rate": "percentage-point difference to same-pitch same-seed PML", "aggregate_asr_cer_delta_pp": "percentage-point difference to same-pitch same-seed PML", "aggregate_asr_wer_delta_pp": "percentage-point difference to same-pitch same-seed PML" }, "pitch_reference": "same variant, same seed, saved native output; not absolute natural F0 truth", "native_pitch_metric": "not_applicable/native_no_edit; no self-identity metric executed", "aggregation": { "aggregation": "Average over the fixed pitches and both seeds within each recording, then equal-weight mean over 5 recordings; balanced condition mean is numerically identical. Descriptive only.", "metric_status_counts": { "automatic_mos": { "available": 90 }, "relative_f0_mae_cents": { "not_applicable": 30, "available": 60 }, "unvoiced_miss_rate": { "not_applicable": 30, "available": 60 }, "recognition": { "unavailable": 54, "available": 36 } }, "metric_reason_counts": { "automatic_mos": {}, "relative_f0_mae_cents": { "native_no_edit": 30 }, "unvoiced_miss_rate": { "native_no_edit": 30 }, "recognition": { "official_excerpt_ground_truth_not_prepared": 54 } }, "raw_mos_recording_macro_means": { "PML": 3.333237099647522, "RAW_SHAPE": 3.397873361905416, "RAW_FULL": 3.4325957934061684 }, "fixed_harm_thresholds": { "automatic_mos": { "direction": "drop", "strict_threshold": 0.15, "unit": "predicted MOS" }, "relative_f0_mae_cents": { "direction": "increase", "strict_threshold": 10, "unit": "cent" }, "unvoiced_miss_rate": { "direction": "increase", "strict_threshold": 2, "unit": "percentage points" } }, "recognition_harm_threshold": null, "global_by_variant": [ { "variant": "RAW_FULL", "recording_count": 5, "all_four_metric_fields_available_conditions": 8, "conditions_with_missing_or_not_applicable_metric_fields": 22, "metrics": { "automatic_mos": { "condition_count": 30, "mean": 0.09935869375864664, "min": -0.12124156951904297, "max": 0.3193175792694092, "negative_count": 6, "zero_count": 0, "positive_count": 24, "recording_macro_mean": 0.09935869375864664, "recording_means_positive": 5, "harm_condition_count": 0, "harm_recording_count": 0 }, "relative_f0_mae_cents": { "condition_count": 20, "mean": 0.47022030297967515, "min": -0.2315186818040047, "max": 4.556269930856767, "negative_count": 10, "zero_count": 0, "positive_count": 10, "recording_macro_mean": 0.4702203029796752, "recording_means_positive": 3, "harm_condition_count": 0, "harm_recording_count": 0 }, "unvoiced_miss_rate": { "condition_count": 20, "mean": -0.36948203865687784, "min": -3.6828737300435406, "max": 1.575622445187662, "negative_count": 13, "zero_count": 2, "positive_count": 5, "recording_macro_mean": -0.36948203865687784, "recording_means_positive": 2, "harm_condition_count": 0, "harm_recording_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 12, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 12, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 12, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 12, "positive_count": 0 } }, { "variant": "RAW_SHAPE", "recording_count": 5, "all_four_metric_fields_available_conditions": 8, "conditions_with_missing_or_not_applicable_metric_fields": 22, "metrics": { "automatic_mos": { "condition_count": 30, "mean": 0.06463626225789389, "min": -0.16630911827087402, "max": 0.2619328498840332, "negative_count": 6, "zero_count": 0, "positive_count": 24, "recording_macro_mean": 0.06463626225789389, "recording_means_positive": 5, "harm_condition_count": 2, "harm_recording_count": 2 }, "relative_f0_mae_cents": { "condition_count": 20, "mean": 0.5038737422636239, "min": -0.22789053881125154, "max": 4.888692673909232, "negative_count": 10, "zero_count": 0, "positive_count": 10, "recording_macro_mean": 0.5038737422636239, "recording_means_positive": 2, "harm_condition_count": 0, "harm_recording_count": 0 }, "unvoiced_miss_rate": { "condition_count": 20, "mean": -0.39131708237900764, "min": -3.6828737300435406, "max": 1.0947204968944106, "negative_count": 13, "zero_count": 1, "positive_count": 6, "recording_macro_mean": -0.39131708237900764, "recording_means_positive": 2, "harm_condition_count": 0, "harm_recording_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 12, "mean": -0.4166666666666667, "min": -5.0, "max": 0.0, "negative_count": 1, "zero_count": 11, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 12, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 12, "positive_count": 0 } } ], "by_recording": [ { "case": "librispeech_1089-134686-0008", "variant": "RAW_FULL", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.1339916785558065, "min": 0.014464139938354492, "max": 0.2585625648498535, "negative_count": 0, "zero_count": 0, "positive_count": 6, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": -0.01924279171863641, "min": -0.04374077498280737, "max": 0.004267486535739273, "negative_count": 3, "zero_count": 0, "positive_count": 1, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": 0.3215638061710255, "min": -0.6736893679568839, "max": 1.5586419753086445, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_1188-133604-0003", "variant": "RAW_FULL", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.13905978202819824, "min": 0.002937793731689453, "max": 0.2781510353088379, "negative_count": 0, "zero_count": 0, "positive_count": 6, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": 0.0385980232310682, "min": -0.10699433996686203, "max": 0.2805916569023075, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -1.1822523980578692, "min": -3.2681604110175533, "max": 0.0, "negative_count": 3, "zero_count": 1, "positive_count": 0, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_121-121726-0012", "variant": "RAW_FULL", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.052343010902404785, "min": -0.12124156951904297, "max": 0.19600343704223633, "negative_count": 2, "zero_count": 0, "positive_count": 4, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": 2.360126317198707, "min": -0.025959485394397408, "max": 4.556269930856767, "negative_count": 1, "zero_count": 0, "positive_count": 3, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": 0.5258268301746558, "min": -0.327015979189893, "max": 1.575622445187662, "negative_count": 1, "zero_count": 1, "positive_count": 2, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 } }, { "case": "librispeech_1221-135766-0005", "variant": "RAW_FULL", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.08638258775075276, "min": -0.05213737487792969, "max": 0.3193175792694092, "negative_count": 2, "zero_count": 0, "positive_count": 4, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": 0.017998166227008783, "min": -0.04292259431989809, "max": 0.05696436118087611, "negative_count": 1, "zero_count": 0, "positive_count": 3, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -0.2850797069399843, "min": -0.29998269330615523, "max": -0.27017672057381426, "negative_count": 4, "zero_count": 0, "positive_count": 0, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_1089-134686-0032", "variant": "RAW_FULL", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.08501640955607097, "min": -0.05608987808227539, "max": 0.18549847602844238, "negative_count": 2, "zero_count": 0, "positive_count": 4, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": -0.046378200039771666, "min": -0.2315186818040047, "max": 0.12505285426511392, "negative_count": 3, "zero_count": 0, "positive_count": 1, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -1.2274687246322171, "min": -3.6828737300435406, "max": 0.47169811320754707, "negative_count": 3, "zero_count": 0, "positive_count": 1, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 } }, { "case": "librispeech_1089-134686-0008", "variant": "RAW_SHAPE", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.07491016387939453, "min": 0.012205839157104492, "max": 0.20004677772521973, "negative_count": 0, "zero_count": 0, "positive_count": 6, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": -0.02011008401947445, "min": -0.07518098879540425, "max": 0.014539253869237712, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": 0.37536277532123596, "min": -0.13882083945778168, "max": 1.0947204968944106, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_1188-133604-0003", "variant": "RAW_SHAPE", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.09798125425974528, "min": 0.03378486633300781, "max": 0.23297953605651855, "negative_count": 0, "zero_count": 0, "positive_count": 6, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": -0.012239285893915097, "min": -0.11884320476810606, "max": 0.10160740234052623, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -0.8326020484075194, "min": -1.869559012416154, "max": 0.0, "negative_count": 3, "zero_count": 1, "positive_count": 0, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_121-121726-0012", "variant": "RAW_SHAPE", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.018649697303771973, "min": -0.15592575073242188, "max": 0.18472695350646973, "negative_count": 3, "zero_count": 0, "positive_count": 3, "harm_condition_count": 1 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": 2.5892217425097144, "min": -0.011997501423662627, "max": 4.888692673909232, "negative_count": 1, "zero_count": 0, "positive_count": 3, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": 0.31215161649944223, "min": -0.8547008547008544, "max": 0.8547008547008544, "negative_count": 1, "zero_count": 0, "positive_count": 3, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 6, "mean": -0.8333333333333334, "min": -5.0, "max": 0.0, "negative_count": 1, "zero_count": 5, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 } }, { "case": "librispeech_1221-135766-0005", "variant": "RAW_SHAPE", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.0708622932434082, "min": -0.029789447784423828, "max": 0.2619328498840332, "negative_count": 1, "zero_count": 0, "positive_count": 5, "harm_condition_count": 0 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": 0.016577167103293156, "min": -0.03548351558617302, "max": 0.09445744831170755, "negative_count": 2, "zero_count": 0, "positive_count": 2, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -0.58402903067598, "min": -0.8554922214827885, "max": -0.2903079579101683, "negative_count": 4, "zero_count": 0, "positive_count": 0, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 0, "mean": null, "min": null, "max": null, "negative_count": 0, "zero_count": 0, "positive_count": 0 } }, { "case": "librispeech_1089-134686-0032", "variant": "RAW_SHAPE", "metrics": { "automatic_mos": { "condition_count": 6, "mean": 0.060777902603149414, "min": -0.16630911827087402, "max": 0.1916975975036621, "negative_count": 2, "zero_count": 0, "positive_count": 4, "harm_condition_count": 1 }, "relative_f0_mae_cents": { "condition_count": 4, "mean": -0.054080828381498525, "min": -0.22789053881125154, "max": 0.07129993408331448, "negative_count": 3, "zero_count": 0, "positive_count": 1, "harm_condition_count": 0 }, "unvoiced_miss_rate": { "condition_count": 4, "mean": -1.2274687246322171, "min": -3.6828737300435406, "max": 0.47169811320754707, "negative_count": 3, "zero_count": 0, "positive_count": 1, "harm_condition_count": 0 } }, "asr_cer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 }, "asr_wer_delta_pp": { "condition_count": 6, "mean": 0.0, "min": 0.0, "max": 0.0, "negative_count": 0, "zero_count": 6, "positive_count": 0 } } ], "flagged_pairs": [ { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "automatic_mos_PML": 3.310621738433838, "automatic_mos_variant": 3.154695987701416, "automatic_mos_delta": -0.15592575073242188, "automatic_mos_harm_flag": true, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": null, "relative_f0_mae_cents_variant": null, "relative_f0_mae_cents_delta": null, "relative_f0_mae_cents_harm_flag": null, "relative_f0_mae_cents_status": "not_applicable", "unvoiced_miss_rate_PML": null, "unvoiced_miss_rate_variant": null, "unvoiced_miss_rate_delta": null, "unvoiced_miss_rate_harm_flag": null, "unvoiced_miss_rate_status": "not_applicable", "expected_voiced_frames_PML": null, "expected_voiced_frames_variant": null, "missing_voiced_frames_PML": null, "missing_voiced_frames_variant": null, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.05, "cer_variant": 0.05, "cer_delta_pp": 0.0, "wer_PML": 0.2, "wer_variant": 0.2, "wer_delta_pp": 0.0, "any_existing_metric_harm": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "automatic_mos_PML": 3.2640607357025146, "automatic_mos_variant": 3.0977516174316406, "automatic_mos_delta": -0.16630911827087402, "automatic_mos_harm_flag": true, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": 3.440849671679332, "relative_f0_mae_cents_variant": 3.5121496057626467, "relative_f0_mae_cents_delta": 0.07129993408331448, "relative_f0_mae_cents_harm_flag": false, "relative_f0_mae_cents_status": "available", "unvoiced_miss_rate_PML": 0.08490566037735849, "unvoiced_miss_rate_variant": 0.04807692307692308, "unvoiced_miss_rate_delta": -3.6828737300435406, "unvoiced_miss_rate_harm_flag": false, "unvoiced_miss_rate_status": "available", "expected_voiced_frames_PML": 106, "expected_voiced_frames_variant": 104, "missing_voiced_frames_PML": 9, "missing_voiced_frames_variant": 5, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.0, "cer_variant": 0.0, "cer_delta_pp": 0.0, "wer_PML": 0.0, "wer_variant": 0.0, "wer_delta_pp": 0.0, "any_existing_metric_harm": true } ], "primary_sample3_plus3": [ { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "automatic_mos_PML": 2.9803714752197266, "automatic_mos_variant": 2.920459747314453, "automatic_mos_delta": -0.05991172790527344, "automatic_mos_harm_flag": false, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": 2.4545880331195735, "relative_f0_mae_cents_variant": 3.2908479408971285, "relative_f0_mae_cents_delta": 0.836259907777555, "relative_f0_mae_cents_harm_flag": false, "relative_f0_mae_cents_status": "available", "unvoiced_miss_rate_PML": 0.08547008547008547, "unvoiced_miss_rate_variant": 0.07692307692307693, "unvoiced_miss_rate_delta": -0.8547008547008544, "unvoiced_miss_rate_harm_flag": false, "unvoiced_miss_rate_status": "available", "expected_voiced_frames_PML": 117, "expected_voiced_frames_variant": 117, "missing_voiced_frames_PML": 10, "missing_voiced_frames_variant": 9, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.05, "cer_variant": 0.05, "cer_delta_pp": 0.0, "wer_PML": 0.2, "wer_variant": 0.2, "wer_delta_pp": 0.0, "any_existing_metric_harm": false }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "automatic_mos_PML": 2.9803714752197266, "automatic_mos_variant": 2.9970176219940186, "automatic_mos_delta": 0.016646146774291992, "automatic_mos_harm_flag": false, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": 2.4545880331195735, "relative_f0_mae_cents_variant": 3.134787593168728, "relative_f0_mae_cents_delta": 0.6801995600491546, "relative_f0_mae_cents_harm_flag": false, "relative_f0_mae_cents_status": "available", "unvoiced_miss_rate_PML": 0.08547008547008547, "unvoiced_miss_rate_variant": 0.08547008547008547, "unvoiced_miss_rate_delta": 0.0, "unvoiced_miss_rate_harm_flag": false, "unvoiced_miss_rate_status": "available", "expected_voiced_frames_PML": 117, "expected_voiced_frames_variant": 117, "missing_voiced_frames_PML": 10, "missing_voiced_frames_variant": 10, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.05, "cer_variant": 0.05, "cer_delta_pp": 0.0, "wer_PML": 0.2, "wer_variant": 0.2, "wer_delta_pp": 0.0, "any_existing_metric_harm": false }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "automatic_mos_PML": 3.065093755722046, "automatic_mos_variant": 3.0451061725616455, "automatic_mos_delta": -0.01998758316040039, "automatic_mos_harm_flag": false, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": 2.495088612212083, "relative_f0_mae_cents_variant": 7.139020501987815, "relative_f0_mae_cents_delta": 4.6439318897757325, "relative_f0_mae_cents_harm_flag": false, "relative_f0_mae_cents_status": "available", "unvoiced_miss_rate_PML": 0.0782608695652174, "unvoiced_miss_rate_variant": 0.08547008547008547, "unvoiced_miss_rate_delta": 0.7209215904868074, "unvoiced_miss_rate_harm_flag": false, "unvoiced_miss_rate_status": "available", "expected_voiced_frames_PML": 115, "expected_voiced_frames_variant": 117, "missing_voiced_frames_PML": 9, "missing_voiced_frames_variant": 10, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.05, "cer_variant": 0.05, "cer_delta_pp": 0.0, "wer_PML": 0.2, "wer_variant": 0.2, "wer_delta_pp": 0.0, "any_existing_metric_harm": false }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "automatic_mos_PML": 3.065093755722046, "automatic_mos_variant": 3.046970844268799, "automatic_mos_delta": -0.01812291145324707, "automatic_mos_harm_flag": false, "automatic_mos_status": "available", "relative_f0_mae_cents_PML": 2.495088612212083, "relative_f0_mae_cents_variant": 6.725083875495386, "relative_f0_mae_cents_delta": 4.229995263283303, "relative_f0_mae_cents_harm_flag": false, "relative_f0_mae_cents_status": "available", "unvoiced_miss_rate_PML": 0.0782608695652174, "unvoiced_miss_rate_variant": 0.09401709401709402, "unvoiced_miss_rate_delta": 1.575622445187662, "unvoiced_miss_rate_harm_flag": false, "unvoiced_miss_rate_status": "available", "expected_voiced_frames_PML": 115, "expected_voiced_frames_variant": 117, "missing_voiced_frames_PML": 9, "missing_voiced_frames_variant": 11, "recognition_status": "available", "recognition_reason": "", "cer_PML": 0.05, "cer_variant": 0.05, "cer_delta_pp": 0.0, "wer_PML": 0.2, "wer_variant": 0.2, "wer_delta_pp": 0.0, "any_existing_metric_harm": false } ], "recognition_original_reference_gate": [ { "case": "librispeech_121-121726-0012", "cer": 0.0, "wer": 0.0, "maximum_reference_cer": 0.2, "passed": true }, { "case": "librispeech_1089-134686-0032", "cer": 0.0, "wer": 0.0, "maximum_reference_cer": 0.2, "passed": true } ], "recognition_partial_count": 0 }, "rows": [ { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.255971908569336, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "66fa9fe6bb6e0880a2213e700ec59f6dd8af48283c5c350d980d948bf12644f6" }, "native_reference": null, "full_metric_record_sha256": "2768f1f8bc70d11364bf977c314bbc7c829d8ced06e4e7215b002beee0bcc70f" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.280392646789551, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b44e98e3c5ac8737a3b10366cda95fcbfec3b5aa87ab5f6c7b2a6bcd99dd8553" }, "native_reference": null, "full_metric_record_sha256": "6ec8206fd104741a2d269769b5a71dbf4eae8242a4fa7bba14a21072e961b9ae" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.313225269317627, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "64a0d355df20f0bc11c6826f22412a88cb5259f79a7f1b115ad836c18fe43376" }, "native_reference": null, "full_metric_record_sha256": "ed357e07ac46c307eec587fdd360799244f8753687333182190f7a70c1f43893" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1469461917877197, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "44c63bf3bc70c9df10e573995949499e4a60667e1836e05b66a3133503156c8d" }, "native_reference": null, "full_metric_record_sha256": "c6d6beb6b0c93abe39fa03bdc17fdc0bd0b5b8c9ab2b337469972f36e445b706" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3469929695129395, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f89b2bdda96a657896c8d1823138a0b4672cbe31281b7f9ad1133ba335d9922c" }, "native_reference": null, "full_metric_record_sha256": "8eeb223ff60a41828d0fdb17e0e40e8b9c5c342773dfa96d48bbe15eb6091691" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4055087566375732, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "558cbee814fcd6343fd7e49491f5030d1d1f10fd111d35b41a9befcd06d87b10" }, "native_reference": null, "full_metric_record_sha256": "481e4c35ec1c65a94997c3c4abc4aa1537e9d33858a03fbfe7d8a0234e40d539" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.5136198958037803, "detail": { "status": "ok", "value": 2.5136198958037803, "unit": "cent", "relative_error": { "count": 122, "mean": 0.17086329166974426, "mae": 2.5136198958037803, "rmse": 3.878035704230505, "abs_p95": 9.486547206370638, "abs_max": 14.307735187790755 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 160, "expected_unvoiced_frames": 433, "missing_voiced_frames": 38, "false_voiced_frames": 0, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2375, "missing_voiced_frames": 38, "expected_voiced_frames": 160, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4511685371398926, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f15ca4b5126702789e289f8d8ec6bd29f659d2cd69f9b5aafabee03a1e1e68ae" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "66fa9fe6bb6e0880a2213e700ec59f6dd8af48283c5c350d980d948bf12644f6" }, "full_metric_record_sha256": "234af39c908cb4487f25efed44306b8ce54973cbe8d292c23de558f589be33f2" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.489294182706981, "detail": { "status": "ok", "value": 2.489294182706981, "unit": "cent", "relative_error": { "count": 121, "mean": 0.12929046779748538, "mae": 2.489294182706981, "rmse": 3.8623766657478438, "abs_p95": 9.566399674662684, "abs_max": 14.308768631290244 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 161, "expected_unvoiced_frames": 432, "missing_voiced_frames": 40, "false_voiced_frames": 0, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2484472049689441, "missing_voiced_frames": 40, "expected_voiced_frames": 161, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.463374376296997, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "372b199b19f5ffcb3e3144eba9e4c1630ca8a9654c8891c42d314958a8269c12" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b44e98e3c5ac8737a3b10366cda95fcbfec3b5aa87ab5f6c7b2a6bcd99dd8553" }, "full_metric_record_sha256": "284cfabd7356760d5516e87245a5c23c030eeec77daf834bb0d5a567d3a7b5f4" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.484885272862165, "detail": { "status": "ok", "value": 2.484885272862165, "unit": "cent", "relative_error": { "count": 121, "mean": 0.12855112998597787, "mae": 2.484885272862165, "rmse": 3.8574578470518617, "abs_p95": 9.557540477544725, "abs_max": 14.308656604444584 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 162, "expected_unvoiced_frames": 431, "missing_voiced_frames": 41, "false_voiced_frames": 0, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25308641975308643, "missing_voiced_frames": 41, "expected_voiced_frames": 162, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5377583503723145, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "36295581701905a6e57570b15f8561b1cc34087a39fd8b45a88419e98085fc96" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "64a0d355df20f0bc11c6826f22412a88cb5259f79a7f1b115ad836c18fe43376" }, "full_metric_record_sha256": "c582d1a1dc72651a9774f52b5409294040574447d69184fa616812befa3b335a" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.510186576403461, "detail": { "status": "ok", "value": 2.510186576403461, "unit": "cent", "relative_error": { "count": 122, "mean": 0.026877113573113964, "mae": 2.510186576403461, "rmse": 3.96711250400346, "abs_p95": 10.717589558371929, "abs_max": 14.302979628642944 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 156, "expected_unvoiced_frames": 437, "missing_voiced_frames": 34, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21794871794871795, "missing_voiced_frames": 34, "expected_voiced_frames": 156, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.558347225189209, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f12964dd85e940bed291b8e5d388d7c148876a334ea7a4413b375c168d80fd43" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "44c63bf3bc70c9df10e573995949499e4a60667e1836e05b66a3133503156c8d" }, "full_metric_record_sha256": "2f4b1cdf9dfcf01c5695bcd979168795b4f297a1a619a865e22419d80cc31261" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.514713688348529, "detail": { "status": "ok", "value": 2.514713688348529, "unit": "cent", "relative_error": { "count": 123, "mean": -0.005282125783168051, "mae": 2.514713688348529, "rmse": 3.9489595764619065, "abs_p95": 10.65108769783211, "abs_max": 14.296582517561017 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 157, "expected_unvoiced_frames": 436, "missing_voiced_frames": 34, "false_voiced_frames": 0, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21656050955414013, "missing_voiced_frames": 34, "expected_voiced_frames": 157, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.639284133911133, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "85a3e6de38bf54142b9957e8b32eab48d645edf50c9aaf839febecc2772d153d" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f89b2bdda96a657896c8d1823138a0b4672cbe31281b7f9ad1133ba335d9922c" }, "full_metric_record_sha256": "fafde2fd081323d3b438aae74580b2b10cbcb11b147d2df3c37d95355f46e132" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.5144540629392003, "detail": { "status": "ok", "value": 2.5144540629392003, "unit": "cent", "relative_error": { "count": 123, "mean": -0.0059870474168069315, "mae": 2.5144540629392003, "rmse": 3.9525631810207837, "abs_p95": 10.649592063160693, "abs_max": 14.297578540668852 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 157, "expected_unvoiced_frames": 436, "missing_voiced_frames": 34, "false_voiced_frames": 0, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.21656050955414013, "missing_voiced_frames": 34, "expected_voiced_frames": 157, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5728113651275635, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "89066009f16527f1236a4f3de5d8d3968d76a6ef510e02ffbb6689d0adfaaf6d" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "558cbee814fcd6343fd7e49491f5030d1d1f10fd111d35b41a9befcd06d87b10" }, "full_metric_record_sha256": "25b01c7623a4e748d5ade2885368f2344ea247b6f14cc10ec725222d2d29ab13" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.9078550878431146, "detail": { "status": "ok", "value": 2.9078550878431146, "unit": "cent", "relative_error": { "count": 150, "mean": -0.07922190408682743, "mae": 2.9078550878431146, "rmse": 4.731753867915315, "abs_p95": 11.21579614022993, "abs_max": 20.431553834918326 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 160, "expected_unvoiced_frames": 433, "missing_voiced_frames": 10, "false_voiced_frames": 24, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0625, "missing_voiced_frames": 10, "expected_voiced_frames": 160, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4076361656188965, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "92bbaffce17b1bd4f39af18f13e049fab44420698441539ebe1084d24a360561" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "66fa9fe6bb6e0880a2213e700ec59f6dd8af48283c5c350d980d948bf12644f6" }, "full_metric_record_sha256": "0a99e7da073d78e67c47d2c874c90326412d09cbea02720943e9cb20a52780ff" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8326740990477104, "detail": { "status": "ok", "value": 2.8326740990477104, "unit": "cent", "relative_error": { "count": 150, "mean": 0.045017610461013505, "mae": 2.8326740990477104, "rmse": 4.578349473124519, "abs_p95": 9.950896824100482, "abs_max": 20.315473481991127 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 161, "expected_unvoiced_frames": 432, "missing_voiced_frames": 11, "false_voiced_frames": 23, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06832298136645963, "missing_voiced_frames": 11, "expected_voiced_frames": 161, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.4383981227874756, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "c22a94e6d2092e0cdba61b2e9a2d6c67826aefaa0ab7a43735a9004dc602d10e" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b44e98e3c5ac8737a3b10366cda95fcbfec3b5aa87ab5f6c7b2a6bcd99dd8553" }, "full_metric_record_sha256": "665d4d7f3f308811161dd36a6479b2a49340cd5928f3d63611fe01184b840384" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8641143128603073, "detail": { "status": "ok", "value": 2.8641143128603073, "unit": "cent", "relative_error": { "count": 151, "mean": 0.07991569460014485, "mae": 2.8641143128603073, "rmse": 4.613646310032524, "abs_p95": 9.951929145753091, "abs_max": 20.297750993054887 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 162, "expected_unvoiced_frames": 431, "missing_voiced_frames": 11, "false_voiced_frames": 22, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06790123456790123, "missing_voiced_frames": 11, "expected_voiced_frames": 162, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5979163646698, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "0010925b8b161a0ce29017235f349eabd0c3dd576a2e1a3b3d1ce38bbce9b5a8" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "64a0d355df20f0bc11c6826f22412a88cb5259f79a7f1b115ad836c18fe43376" }, "full_metric_record_sha256": "074f78a92707607c40fbe083e08ad14727a06bfdf94386dcb060a82055b9efe6" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.878742884045959, "detail": { "status": "ok", "value": 2.878742884045959, "unit": "cent", "relative_error": { "count": 147, "mean": -0.08714220818404807, "mae": 2.878742884045959, "rmse": 4.679829196651522, "abs_p95": 11.37575549328304, "abs_max": 21.760517628447722 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 156, "expected_unvoiced_frames": 437, "missing_voiced_frames": 9, "false_voiced_frames": 25, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.057692307692307696, "missing_voiced_frames": 9, "expected_voiced_frames": 156, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0997512340545654, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "dd6507d9d930d7aa7d7f1c3fde3838d460cb15162848328f52a9e2abd0f801fc" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "44c63bf3bc70c9df10e573995949499e4a60667e1836e05b66a3133503156c8d" }, "full_metric_record_sha256": "27a61ba8c3f1fe948a1e0350407e94aae34ac9ca898eb6f853b52d46638f986e" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8932821379151967, "detail": { "status": "ok", "value": 2.8932821379151967, "unit": "cent", "relative_error": { "count": 148, "mean": -0.1115351559646525, "mae": 2.8932821379151967, "rmse": 4.721626097787265, "abs_p95": 11.589163743568443, "abs_max": 21.860337983067986 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 157, "expected_unvoiced_frames": 436, "missing_voiced_frames": 9, "false_voiced_frames": 23, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.05732484076433121, "missing_voiced_frames": 9, "expected_voiced_frames": 157, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2008399963378906, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f7abbf66839bd3897b5719d9e2fd9547e51a1aa9dae96809bcc7c595f7a963ee" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "f89b2bdda96a657896c8d1823138a0b4672cbe31281b7f9ad1133ba335d9922c" }, "full_metric_record_sha256": "9b06f69e349dbad0bd405fa835e0614d51c712a1fa8867bf6ba90be9d3767d8f" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.8699796285600967, "detail": { "status": "ok", "value": 2.8699796285600967, "unit": "cent", "relative_error": { "count": 149, "mean": -0.10059026913747586, "mae": 2.8699796285600967, "rmse": 4.696618283694809, "abs_p95": 11.467407799477828, "abs_max": 21.836592528378745 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 157, "expected_unvoiced_frames": 436, "missing_voiced_frames": 8, "false_voiced_frames": 24, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.050955414012738856, "missing_voiced_frames": 8, "expected_voiced_frames": 157, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.29655122756958, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 107680, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "331f6cc53fe945354bc2a0e36c458c4eb6ec5e931c4b95131f4659ca52938ebf" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "558cbee814fcd6343fd7e49491f5030d1d1f10fd111d35b41a9befcd06d87b10" }, "full_metric_record_sha256": "4221b6d4ef178bb7c80f9280920d2d15b6378315c1de158df73b94f83371e20b" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.425837993621826, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "d3e29b6ecedbb29f259d2b5da2790a9fb769b78cda50fe1daae75d47da0e9713" }, "native_reference": null, "full_metric_record_sha256": "df7eaf1152f42677ba7671556d228d18e248e15f63d36cb8c50f001c7d670062" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5048186779022217, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b5986f69364cea1af9ca0584d106968a708d0303cff0b3a963236f8a1050ddf1" }, "native_reference": null, "full_metric_record_sha256": "a87ec44484afc22d079ccb6aae5c8844162bf84f5d80f523c5c402f34a754000" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.509075880050659, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "2b1046ba58dad230b4eb3350850b571682f9f862c5cb9f88b6ea1863ade21b3b" }, "native_reference": null, "full_metric_record_sha256": "d208f3035c5eb1baf59ba6626f50c476a12d54d8926a228458843258621fab99" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.123307466506958, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "ce73d360c6561f425030780b9bb41936ba0db8217858c516d53dc30f465c65c5" }, "native_reference": null, "full_metric_record_sha256": "dfb8a3b468a366038a97d37e7495b659c8697e09416423c2da64dc06e6fbdac2" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2155163288116455, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "a015d2fcc1eec56bfa3722ede0fb22d357d17264b5bbdaf28a79d04b866edec0" }, "native_reference": null, "full_metric_record_sha256": "78cde8ff826ff7a14b675884b9540da5e17bbb38ee3f1793ecf8190ace277a13" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.298588752746582, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "69b6cc09a19df04c7917894c44a2e69314ec8cb40ea0cd5a6a6cf505992bd07c" }, "native_reference": null, "full_metric_record_sha256": "50c0bcd3057c15bf9869b1fcf9c78d307748d0ce8393290d546f9bccb6f3f312" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2146110975973308, "detail": { "status": "ok", "value": 2.2146110975973308, "unit": "cent", "relative_error": { "count": 110, "mean": 0.38958306141279847, "mae": 2.2146110975973308, "rmse": 3.1796525073663413, "abs_p95": 5.676959927118201, "abs_max": 14.105843456441928 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 31, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "missing_voiced_frames": 31, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5976874828338623, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "1bc1d3429be451315e5be2fe6196e4bd6667a0145169ab797717f13c9adc67ca" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "d3e29b6ecedbb29f259d2b5da2790a9fb769b78cda50fe1daae75d47da0e9713" }, "full_metric_record_sha256": "5f6e7ff7ec69a1b4b62d1454d0dba03019ffbfdfd12eef79aca384e89b8d5982" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.167878958269385, "detail": { "status": "ok", "value": 2.167878958269385, "unit": "cent", "relative_error": { "count": 110, "mean": 0.37340669142412997, "mae": 2.167878958269385, "rmse": 3.1515893241933073, "abs_p95": 5.643802239026566, "abs_max": 14.292976552234034 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 31, "false_voiced_frames": 13, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "missing_voiced_frames": 31, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7116270065307617, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "feb1ee37fd381b6aba4acae0489a8c7079ba7dd637745b8c055d541bac4468e3" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b5986f69364cea1af9ca0584d106968a708d0303cff0b3a963236f8a1050ddf1" }, "full_metric_record_sha256": "3f66c3ad7ea0378e04e21b70aea4ae3b6f7c17046833f80515024019f6a02ac8" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.1716774994939865, "detail": { "status": "ok", "value": 2.1716774994939865, "unit": "cent", "relative_error": { "count": 110, "mean": 0.3561695425011477, "mae": 2.1716774994939865, "rmse": 3.1519603747821257, "abs_p95": 5.637871494821099, "abs_max": 14.344185642064348 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 31, "false_voiced_frames": 13, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2198581560283688, "missing_voiced_frames": 31, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.754973888397217, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "e9e7cf6d978ef93e71fc7009d1de5523516949eb8174f493ee618e3d6073d6bb" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "2b1046ba58dad230b4eb3350850b571682f9f862c5cb9f88b6ea1863ade21b3b" }, "full_metric_record_sha256": "57e23717da0d7c7978abc33ca099bd3cbcccaca943042ad67af7727c3c065bc2" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.399544012376205, "detail": { "status": "ok", "value": 2.399544012376205, "unit": "cent", "relative_error": { "count": 113, "mean": 0.011635035144058283, "mae": 2.399544012376205, "rmse": 3.801861266569337, "abs_p95": 6.661569671395328, "abs_max": 21.277909871927854 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 147, "expected_unvoiced_frames": 446, "missing_voiced_frames": 34, "false_voiced_frames": 4, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.23129251700680273, "missing_voiced_frames": 34, "expected_voiced_frames": 147, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5601258277893066, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "4f69304610d511311c2936f8d8d81b7737e0de2128026f1656e4cc4a57a02522" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "ce73d360c6561f425030780b9bb41936ba0db8217858c516d53dc30f465c65c5" }, "full_metric_record_sha256": "881c099975424b2359a4f66dd4d5981af0a7eab747a9c1a7d8a6a5adb4876118" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.41455481055607, "detail": { "status": "ok", "value": 2.41455481055607, "unit": "cent", "relative_error": { "count": 111, "mean": 0.11049391349078198, "mae": 2.41455481055607, "rmse": 3.907679506085886, "abs_p95": 6.323368470753596, "abs_max": 25.20377646451675 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 143, "expected_unvoiced_frames": 450, "missing_voiced_frames": 32, "false_voiced_frames": 6, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "missing_voiced_frames": 32, "expected_voiced_frames": 143, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.793105363845825, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "d91225ad0a3d20e676e120b58fc3bfa251c488d95fc031dd52b70cd2682275ce" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "a015d2fcc1eec56bfa3722ede0fb22d357d17264b5bbdaf28a79d04b866edec0" }, "full_metric_record_sha256": "08c30fdf1c58796b3698454af4d617bf82e69d8ccb07a7b72368d482f018f3c5" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4232723864683767, "detail": { "status": "ok", "value": 2.4232723864683767, "unit": "cent", "relative_error": { "count": 111, "mean": 0.11808683518516172, "mae": 2.4232723864683767, "rmse": 3.887112698806267, "abs_p95": 6.304539241379302, "abs_max": 24.770679570810216 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 143, "expected_unvoiced_frames": 450, "missing_voiced_frames": 32, "false_voiced_frames": 6, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.22377622377622378, "missing_voiced_frames": 32, "expected_voiced_frames": 143, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8382768630981445, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "44010ecd3e30fea37fc314d75707781e1394fc92a8acbdf76317d19ddbb27257" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "69b6cc09a19df04c7917894c44a2e69314ec8cb40ea0cd5a6a6cf505992bd07c" }, "full_metric_record_sha256": "a5ee77a0ffddce82b38982677cefb11c2dca7820748b1c41df5aa86482da8605" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.159110101218769, "detail": { "status": "ok", "value": 3.159110101218769, "unit": "cent", "relative_error": { "count": 135, "mean": 0.8250385622762711, "mae": 3.159110101218769, "rmse": 5.003808179652222, "abs_p95": 9.326924761019075, "abs_max": 27.910774856774953 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 6, "false_voiced_frames": 17, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0425531914893617, "missing_voiced_frames": 6, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1961922645568848, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "65085023011d3842b8018ac786d6928eb5668e00256ac74ce689082a6751bae7" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "d3e29b6ecedbb29f259d2b5da2790a9fb769b78cda50fe1daae75d47da0e9713" }, "full_metric_record_sha256": "375297503f1db66bf0807b1d39912cac9e6c67f0c64465058abdd498d6eda6a3" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.040266896450663, "detail": { "status": "ok", "value": 3.040266896450663, "unit": "cent", "relative_error": { "count": 136, "mean": 0.5722304950577933, "mae": 3.040266896450663, "rmse": 4.991739341554132, "abs_p95": 10.316147619110628, "abs_max": 27.893082895788478 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 5, "false_voiced_frames": 16, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "missing_voiced_frames": 5, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2299771308898926, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "7f6495887d5f5d47e2cbd5728681bcf0343669b8ba4f5f086823412ee6e031c8" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b5986f69364cea1af9ca0584d106968a708d0303cff0b3a963236f8a1050ddf1" }, "full_metric_record_sha256": "2a503f25d5c5818e0e47d13871a523b6d3cba17b915bd80e200f160f060e7cb8" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.052115761251907, "detail": { "status": "ok", "value": 3.052115761251907, "unit": "cent", "relative_error": { "count": 136, "mean": 0.5685647949137644, "mae": 3.052115761251907, "rmse": 5.037260436693963, "abs_p95": 10.399707479053465, "abs_max": 27.76638432928407 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 141, "expected_unvoiced_frames": 452, "missing_voiced_frames": 5, "false_voiced_frames": 16, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03546099290780142, "missing_voiced_frames": 5, "expected_voiced_frames": 141, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3336565494537354, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "c6c06a298030eca3f7adb7fdf9851e5323bd99bf8813b2cc79fccfaaa810569f" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "2b1046ba58dad230b4eb3350850b571682f9f862c5cb9f88b6ea1863ade21b3b" }, "full_metric_record_sha256": "9ecaa14f6ac9b885aa53f748b5edfd3521094d214c83d7b2eb2e1bb280e09795" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.993947939082237, "detail": { "status": "ok", "value": 2.993947939082237, "unit": "cent", "relative_error": { "count": 135, "mean": 0.6691744813436629, "mae": 2.993947939082237, "rmse": 4.9111858771283865, "abs_p95": 10.554272353896328, "abs_max": 27.854487528081048 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 147, "expected_unvoiced_frames": 446, "missing_voiced_frames": 12, "false_voiced_frames": 14, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08163265306122448, "missing_voiced_frames": 12, "expected_voiced_frames": 147, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.144958972930908, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "09f09ae312805ea0f54ca68227aea0ca8c306e6f44376919ecd16456b1289267" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "ce73d360c6561f425030780b9bb41936ba0db8217858c516d53dc30f465c65c5" }, "full_metric_record_sha256": "e733b0144d3989a9238c4165a992fbc4cf41c1022553fd24b6644ca4965f6295" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0955553414227635, "detail": { "status": "ok", "value": 3.0955553414227635, "unit": "cent", "relative_error": { "count": 134, "mean": 0.7303747867917804, "mae": 3.0955553414227635, "rmse": 5.043545854358552, "abs_p95": 10.605235781637473, "abs_max": 27.761304965643003 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 143, "expected_unvoiced_frames": 450, "missing_voiced_frames": 9, "false_voiced_frames": 15, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.06293706293706294, "missing_voiced_frames": 9, "expected_voiced_frames": 143, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.180953025817871, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "653056c6e110bfae7b79eafd8b68096b60900b42b7ea51adf633d36242001a34" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "a015d2fcc1eec56bfa3722ede0fb22d357d17264b5bbdaf28a79d04b866edec0" }, "full_metric_record_sha256": "0d6461dd1bb923837ffee33e17cd7e38c805880b1163830699a355167cb4ba42" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.2745395959845447, "detail": { "status": "ok", "value": 3.2745395959845447, "unit": "cent", "relative_error": { "count": 136, "mean": 0.5930446058284259, "mae": 3.2745395959845447, "rmse": 5.409642289444568, "abs_p95": 11.024055167283903, "abs_max": 27.743546431109557 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 143, "expected_unvoiced_frames": 450, "missing_voiced_frames": 7, "false_voiced_frames": 15, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04895104895104895, "missing_voiced_frames": 7, "expected_voiced_frames": 143, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.1478967666625977, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 201760, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "9a89cff025a29917a37ef97f4c3544043d2d5ecb3070ecaceadfeac91bf9435d" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "69b6cc09a19df04c7917894c44a2e69314ec8cb40ea0cd5a6a6cf505992bd07c" }, "full_metric_record_sha256": "09ee7bf57d8edafe3f74c74e09f01cbfb88a894d58d788481634a51fe3cf9a27" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3045575618743896, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "9ee33e50077ddd862c5c71be024d083b3b38045e013b39ec8331d19f1df81d12" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "bee6b53a7fe774160c03a93083e906a9b38cf640e006db02ed7ff8660b282ace" }, "native_reference": null, "full_metric_record_sha256": "f49a8867dd2884693683d1bcf56ddd7ebe3117a40ebc7842eaeb5e70bab09fef" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3706469535827637, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "e1c0ce06444819e8eb623fc85604655ca1ec738625575946072842a9db5937f8" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "b39de8c9555ba9c5ece7bf31410698367738566f4162ba15b4a1b1ff746d4c62" }, "native_reference": null, "full_metric_record_sha256": "00915e297e444eeda23645c23a479015f26e441459f7caaf0f55208e12c74a48" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3967278003692627, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "1e7ace43ec3c647ac781f46d14a5c1958c1145e4326653263cc9ef4eb7c00859" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "6cf3876756c3750b646ec255f2594c3fa5e5030e632c9bbeb1c8b015d9d660d4" }, "native_reference": null, "full_metric_record_sha256": "1fcc9bc5ab9a61a0ea6e2f9e8a76a414593f97c80da2697f5ed1be3e4d516a85" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.310621738433838, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "a93c8beb05c5466cf43a7e269148c82562e96c0b03956c43f2e2ea75205a3217" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "3c80cb0183665bde7489d084712b09a3f802d74c5f31cb9ac1ea25c0dd609bb9" }, "native_reference": null, "full_metric_record_sha256": "f9f67739fb9d8f59b0ac4efaece9db5e632ba60d79a5f98017305a07ac0fb852" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.154695987701416, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "ce820c0cd756295a76dc43c3641dc9870fa9f992b18ad85e3c907337bf4c97b8" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "9e304a730282597644a2cb98f8463de959fe67e82ec538ac95b24713eadaae41" }, "native_reference": null, "full_metric_record_sha256": "41f9759dba17ffe58818e6a9faf30e85982e4e37d69c2b10c021072bcf6fe963" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.189380168914795, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "ba228b71c75d8c4f573b42cec51ae9cc2749abcdbce61adb1880eb10b496f262" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "23f5ecac6f0600c8b612ec2e5464efb96be2ced7d9e8d316af0ada76e8e46f93" }, "native_reference": null, "full_metric_record_sha256": "8883786c03a45c9ff29d46e280bee064becae8e3fd6ae2fa0687d39c0eb5ccfc" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.660224749853958, "detail": { "status": "ok", "value": 2.660224749853958, "unit": "cent", "relative_error": { "count": 95, "mean": -0.412986157931136, "mae": 2.660224749853958, "rmse": 5.6608317214944055, "abs_p95": 8.014136544779312, "abs_max": 37.679788398254885 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 22, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "missing_voiced_frames": 22, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.854731798171997, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "abbd58c0aacccb2c72c32afae2daf4dca5dd067db596fd24080add4fcef42cf8" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "b5d58d1a3452c7cced18f708f3b23c109acb016e90f0ab159606d447fa2a8ba2" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "bee6b53a7fe774160c03a93083e906a9b38cf640e006db02ed7ff8660b282ace" }, "full_metric_record_sha256": "b6764c815672284848dfb610efcf5215f5ddd712b0ef14ea8cbee097b67280a3" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6482272484302953, "detail": { "status": "ok", "value": 2.6482272484302953, "unit": "cent", "relative_error": { "count": 94, "mean": -0.5188366691394211, "mae": 2.6482272484302953, "rmse": 5.775288805157294, "abs_p95": 9.921545656171055, "abs_max": 40.16925809930113 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 23, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "missing_voiced_frames": 23, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.951638698577881, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.1, "detail": { "status": "ok", "value": 0.1, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 2, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.1, "delta": 0.1 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "37f903010c62bd57df9d5ed965e267ed507f160ff323a09c2966e0cd31d4120a" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "758b0ed995d72ee4715440ce181c32f90ce018986d8010890368718a09bae037" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "b39de8c9555ba9c5ece7bf31410698367738566f4162ba15b4a1b1ff746d4c62" }, "full_metric_record_sha256": "35403441b7d5823fe95683fa2c2a18327a71df954e09194ff9a56df8d33c7c3b" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6342652644595606, "detail": { "status": "ok", "value": 2.6342652644595606, "unit": "cent", "relative_error": { "count": 94, "mean": -0.5277803199718009, "mae": 2.6342652644595606, "rmse": 5.764163311422851, "abs_p95": 9.912892818593875, "abs_max": 39.8129042146706 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 23, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "missing_voiced_frames": 23, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0033345222473145, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.15, "detail": { "status": "ok", "value": 0.15, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 3, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.15, "delta": 0.15 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "f66721fa6b6cc1c8b7a816012c682fcec061b261b93a2913358321f41cb21f00" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "c8da7838eee5e846ea1ea3e638a4f7270f7be433fc9c893642421fd34bbe512e" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "6cf3876756c3750b646ec255f2594c3fa5e5030e632c9bbeb1c8b015d9d660d4" }, "full_metric_record_sha256": "281e2df940e9c2d9acffff3a700de51ea39a2544f96458c644c285d018090901" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4626724244784137, "detail": { "status": "ok", "value": 2.4626724244784137, "unit": "cent", "relative_error": { "count": 93, "mean": 0.20134367825813423, "mae": 2.4626724244784137, "rmse": 4.890082080063057, "abs_p95": 7.0943858650431695, "abs_max": 28.396090774519845 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 115, "expected_unvoiced_frames": 282, "missing_voiced_frames": 22, "false_voiced_frames": 4, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19130434782608696, "missing_voiced_frames": 22, "expected_voiced_frames": 115, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.5711166858673096, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "8b246c072b975605df0b2d5059b427f30aeb332e68576e9a1abeccce128c4e79" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "2ed85deb832fe249f779569e7c6379b57d3db0525df04a26444f4d5151db0b89" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "3c80cb0183665bde7489d084712b09a3f802d74c5f31cb9ac1ea25c0dd609bb9" }, "full_metric_record_sha256": "b48c6181b87d007016ec63cced5ec1b1dd5921a0f288a7383e755953569e7193" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.351365098387646, "detail": { "status": "ok", "value": 7.351365098387646, "unit": "cent", "relative_error": { "count": 94, "mean": 4.914736606141206, "mae": 7.351365098387646, "rmse": 36.15264051457129, "abs_p95": 16.180512238105386, "abs_max": 325.8747017742423 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 23, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.19658119658119658, "missing_voiced_frames": 23, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.7558436393737793, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "99fa71763f99e1d85fc54d9407b584e8932448fb23be84ae5d618cffef019351" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "efd7cd6a2dca10baebb8f1cc44f3bf043cb8e0ead5c5927beb0d0ccf208f03f7" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "9e304a730282597644a2cb98f8463de959fe67e82ec538ac95b24713eadaae41" }, "full_metric_record_sha256": "911cef3626443d11ca85a08a74961896dbf05acb49284f4cea7cfa713bc9b665" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.018942355335181, "detail": { "status": "ok", "value": 7.018942355335181, "unit": "cent", "relative_error": { "count": 95, "mean": 4.586803206634048, "mae": 7.018942355335181, "rmse": 34.917694641109826, "abs_p95": 16.142461133987318, "abs_max": 321.03471661199 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 22, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.18803418803418803, "missing_voiced_frames": 22, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.767120122909546, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "332345eab61ad76189c279a1174d55972d52393d24f000ebf2989e6397ca3a4d" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "5301869c730597ada086b2a0b7274654e22a93872364e3a7868e1d74098902c2" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "23f5ecac6f0600c8b612ec2e5464efb96be2ced7d9e8d316af0ada76e8e46f93" }, "full_metric_record_sha256": "d8ecd3132960a2da8a7d28978996eb72ba54df091e619fea2dd845af53788633" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.4545880331195735, "detail": { "status": "ok", "value": 2.4545880331195735, "unit": "cent", "relative_error": { "count": 107, "mean": -0.4110847027229457, "mae": 2.4545880331195735, "rmse": 4.017541854064593, "abs_p95": 7.553790132963368, "abs_max": 23.143858491634504 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 10, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "missing_voiced_frames": 10, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9803714752197266, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "880220051500e7a7618271bba36d9e876c4d9530f64c1b6a30c52bd55c13c7d7" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "082bd02b2d99751a844f503d91a3c6df5c7f98d60aef01c2c8d9ee481295c776" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "bee6b53a7fe774160c03a93083e906a9b38cf640e006db02ed7ff8660b282ace" }, "full_metric_record_sha256": "bc59ef5e94c72f56447a4eba582a443615d71e2d65a6859b2e5c3fcf5c13e46c" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.2908479408971285, "detail": { "status": "ok", "value": 3.2908479408971285, "unit": "cent", "relative_error": { "count": 108, "mean": -1.359601440303928, "mae": 3.2908479408971285, "rmse": 8.953062051517795, "abs_p95": 8.260224822659037, "abs_max": 82.67537642399601 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 9, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.07692307692307693, "missing_voiced_frames": 9, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.920459747314453, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "72a51c6d256965840b9e92b40125e0380b6f75d6a2216e63dee7414a5f5fba0d" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "1ee658832e1ddbc0fead3358ce91f9abb4c6243747772c1b1995ed0390499693" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "b39de8c9555ba9c5ece7bf31410698367738566f4162ba15b4a1b1ff746d4c62" }, "full_metric_record_sha256": "6a3951d87304fe461c1dc9a901a0b3404f9de15e51fcff336bddd7264af2ddf0" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.134787593168728, "detail": { "status": "ok", "value": 3.134787593168728, "unit": "cent", "relative_error": { "count": 107, "mean": -1.1799334798540064, "mae": 3.134787593168728, "rmse": 8.849220592639966, "abs_p95": 8.476268433965602, "abs_max": 82.39683835407061 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 10, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "missing_voiced_frames": 10, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 2.9970176219940186, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "6b7e3b899224966739c2be407162eb93b4db49e609430a8f9948b1f9361b1bb5" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "448a41231e539e37d40953699caf50f78ed0866197fa9c0aa01c689ca773f9f8" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "6cf3876756c3750b646ec255f2594c3fa5e5030e632c9bbeb1c8b015d9d660d4" }, "full_metric_record_sha256": "16ab904b2bc1488b239a7d3dda5269b9a665a500bdb3f39a768cb1de50bcf51f" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.495088612212083, "detail": { "status": "ok", "value": 2.495088612212083, "unit": "cent", "relative_error": { "count": 106, "mean": -0.030031645825742696, "mae": 2.495088612212083, "rmse": 4.613520434449651, "abs_p95": 8.187992079974167, "abs_max": 25.824108340419443 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 115, "expected_unvoiced_frames": 282, "missing_voiced_frames": 9, "false_voiced_frames": 13, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0782608695652174, "missing_voiced_frames": 9, "expected_voiced_frames": 115, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.065093755722046, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "ed7dde62116f60e9bb7337a864266e57362101cd0fade25bd0f5210c1d0434a1" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "d56d4907bb8ee291d7ca7ff2c240ef4a1bdb18d485c78449b45dd152818c0e26" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "3c80cb0183665bde7489d084712b09a3f802d74c5f31cb9ac1ea25c0dd609bb9" }, "full_metric_record_sha256": "fea4fc9d629d8103302210bba0171f5825db3209d4c4e5cdef4ec1ee26b591a0" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 7.139020501987815, "detail": { "status": "ok", "value": 7.139020501987815, "unit": "cent", "relative_error": { "count": 107, "mean": -4.946845390957498, "mae": 7.139020501987815, "rmse": 49.17749852401295, "abs_p95": 8.298794228658315, "abs_max": 506.3110805389724 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 10, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08547008547008547, "missing_voiced_frames": 10, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0451061725616455, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "004f2e5987628e988c09ae5d12c918a2e230837421b542ed84296df62c061b26" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "4f8dc9a31dba23758b8ba40719eee80f25a3ac6ad731fa319bf29302288eac6c" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "9e304a730282597644a2cb98f8463de959fe67e82ec538ac95b24713eadaae41" }, "full_metric_record_sha256": "2a8a21c42a6b1df4d10dbdec5bd4b0428a8fd5653f2569e2086ff6f76ea47bfd" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 6.725083875495386, "detail": { "status": "ok", "value": 6.725083875495386, "unit": "cent", "relative_error": { "count": 106, "mean": -4.701481143781234, "mae": 6.725083875495386, "rmse": 45.15726076453911, "abs_p95": 8.30536998681295, "abs_max": 462.34904966061083 }, "absolute_target_error": null, "total_frames": 397, "expected_voiced_frames": 117, "expected_unvoiced_frames": 280, "missing_voiced_frames": 11, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.09401709401709402, "missing_voiced_frames": 11, "expected_voiced_frames": 117, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.046970844268799, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.05, "detail": { "status": "ok", "value": 0.05, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 20, "reference_rate": 0.0, "output_rate": 0.05, "delta": 0.05 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 1, "truth_tokens": 5, "reference_rate": 0.0, "output_rate": 0.2, "delta": 0.2 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "2c315d2b8330878bdd5d6621ccbf826f5f5b5a57f617e53f8888699c22449e68" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "53a9de0adae6d1f1b59307c0e9d5345f9cbdd7fc155be3ba20e9327d5ae7085b" }, "native_reference": { "bytes": 517888, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 64720, "sha256": "23f5ecac6f0600c8b612ec2e5464efb96be2ced7d9e8d316af0ada76e8e46f93" }, "full_metric_record_sha256": "993cd953a30ef7728e3f185f2d3e0f4010d9669ad356010b4c99550c0b56d4f5" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.517606496810913, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "82cfae761e5c0f68551f01def9e993c41545f7fc8d731b60e1305811d7f63067" }, "native_reference": null, "full_metric_record_sha256": "a56d20523d715be285e487103b768b63d3cc1e1ff76b7787d33c5b967cb7a4d8" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.554678440093994, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "546be5acbc36a76c03db15837211519607b7c74777715eba9bb6794184b556ca" }, "native_reference": null, "full_metric_record_sha256": "c98e3fdf2cbdaaa3c95f209d1be479bb0c2bb1993170128bb4e51d62975cd0f2" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.646789073944092, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b45a6380b5c8f17818af0dfae734ac4982a0ee6a4490f8dfa38759fe2f4e2cb5" }, "native_reference": null, "full_metric_record_sha256": "7773c9e8cbebacf4cad2ffde9d200b4e66a53ef28563af2b098491a9949789dd" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3315794467926025, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "dd0c86f3d424bc48154feff1e81ed6de6d4108469e419c8fcb13f5642277f77c" }, "native_reference": null, "full_metric_record_sha256": "473a9dac8f92425ea5456f192e6cfb229ac211e2ebe075d4e37db3bca3b0c0c4" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3542239665985107, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "e0752efd9d526392bf75afa41d5a3032ad67731c86a028a46fa73f2566f68aa2" }, "native_reference": null, "full_metric_record_sha256": "3318fd9dcd34b0a5710592c4be62be085e9c14b1b0e651c600625ef82d1fd5de" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2800745964050293, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "4b45a182c576a67cdcab118faccba5276fdf67db1ce2539efa7050809fd74bce" }, "native_reference": null, "full_metric_record_sha256": "98dd022df5af45c80e4fea3e7f04b14795e59bb10b2d67b69bc9aa479efac0c7" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.52703824130881, "detail": { "status": "ok", "value": 2.52703824130881, "unit": "cent", "relative_error": { "count": 281, "mean": -0.4190706308697399, "mae": 2.52703824130881, "rmse": 4.095673274734783, "abs_p95": 9.908089960591454, "abs_max": 22.98218149332397 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 323, "expected_unvoiced_frames": 270, "missing_voiced_frames": 42, "false_voiced_frames": 2, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.13003095975232198, "missing_voiced_frames": 42, "expected_voiced_frames": 323, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8351967334747314, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "8fcc061d8e5ae305f5549bbc754d3df844c665be5f6e92cda03ad02b1c838b63" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "82cfae761e5c0f68551f01def9e993c41545f7fc8d731b60e1305811d7f63067" }, "full_metric_record_sha256": "0d6465bf19918572e3d8e61e7213d971760ccb1f1e556cac0098d7464cba4de1" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.6214956896205175, "detail": { "status": "ok", "value": 2.6214956896205175, "unit": "cent", "relative_error": { "count": 282, "mean": -0.5361963659869124, "mae": 2.6214956896205175, "rmse": 4.443430812625779, "abs_p95": 10.70876625998881, "abs_max": 28.202640598013847 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 40, "false_voiced_frames": 2, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12422360248447205, "missing_voiced_frames": 40, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8054072856903076, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "4f9b1ae812443c7b4f5996d20cf2dca1b5622345b8ec3600e104661997ce80ab" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "546be5acbc36a76c03db15837211519607b7c74777715eba9bb6794184b556ca" }, "full_metric_record_sha256": "3a29c27574744087fbfa8ecb11707d1bb239a3d86af8a6dcda416fd82edced16" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.584002602489686, "detail": { "status": "ok", "value": 2.584002602489686, "unit": "cent", "relative_error": { "count": 281, "mean": -0.5031678291716365, "mae": 2.584002602489686, "rmse": 4.363785936278644, "abs_p95": 10.747445112803081, "abs_max": 26.65701175232302 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 41, "false_voiced_frames": 2, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12732919254658384, "missing_voiced_frames": 41, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7830593585968018, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "28d85162fd6360c67380745db8d1e999c220ab137c620b9e38a89f9deab98eff" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b45a6380b5c8f17818af0dfae734ac4982a0ee6a4490f8dfa38759fe2f4e2cb5" }, "full_metric_record_sha256": "00bd830c69daf9e6ad871b0c784586fb5e661e8e74e5ce7ff3a9d966648fc63e" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.661235166123176, "detail": { "status": "ok", "value": 2.661235166123176, "unit": "cent", "relative_error": { "count": 282, "mean": -0.48200064833016565, "mae": 2.661235166123176, "rmse": 4.5574224424585, "abs_p95": 10.669156591389271, "abs_max": 27.74681904284523 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 323, "expected_unvoiced_frames": 270, "missing_voiced_frames": 41, "false_voiced_frames": 1, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12693498452012383, "missing_voiced_frames": 41, "expected_voiced_frames": 323, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.8320510387420654, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b28b91f45e924e38cd9dab10984e28bb2421ad7945e0fefeddb5e1730d69ad26" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "dd0c86f3d424bc48154feff1e81ed6de6d4108469e419c8fcb13f5642277f77c" }, "full_metric_record_sha256": "7295f0b84cef807d039c4a2f4580a7b744b141223c292071cffc577e80c56a0b" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.625751650537003, "detail": { "status": "ok", "value": 2.625751650537003, "unit": "cent", "relative_error": { "count": 283, "mean": -0.5372795871839846, "mae": 2.625751650537003, "rmse": 4.4808404082716065, "abs_p95": 10.511215969187942, "abs_max": 27.963181119120097 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 321, "expected_unvoiced_frames": 272, "missing_voiced_frames": 38, "false_voiced_frames": 2, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.11838006230529595, "missing_voiced_frames": 38, "expected_voiced_frames": 321, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.846437931060791, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "9c55bd5513136d74c0632debc3da1389eaab09c5b8c6e2eeb967247bc652979b" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "e0752efd9d526392bf75afa41d5a3032ad67731c86a028a46fa73f2566f68aa2" }, "full_metric_record_sha256": "7e57b8682d9d8fc295dcea06e834c31cc24ed8ada9ef300d4e4f17e8533f1d2c" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.618312571803278, "detail": { "status": "ok", "value": 2.618312571803278, "unit": "cent", "relative_error": { "count": 282, "mean": -0.5080640176035538, "mae": 2.618312571803278, "rmse": 4.476895072700993, "abs_p95": 10.452941671381092, "abs_max": 26.876287519921576 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 40, "false_voiced_frames": 2, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.12422360248447205, "missing_voiced_frames": 40, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.900097131729126, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "7c14268757b7b8b63cbd5e3a648036dadd57a8ed84b404b863dae6fb37a75e93" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "4b45a182c576a67cdcab118faccba5276fdf67db1ce2539efa7050809fd74bce" }, "full_metric_record_sha256": "aec742504fd7994bf5e5c5b6b039dfae45d1d1968dc46b45d9390a21759214a7" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.108724576950197, "detail": { "status": "ok", "value": 3.108724576950197, "unit": "cent", "relative_error": { "count": 311, "mean": -1.6046185353085438, "mae": 3.108724576950197, "rmse": 5.303751809966267, "abs_p95": 11.596624410479878, "abs_max": 39.37342245100803 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 323, "expected_unvoiced_frames": 270, "missing_voiced_frames": 12, "false_voiced_frames": 27, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03715170278637771, "missing_voiced_frames": 12, "expected_voiced_frames": 323, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.332209825515747, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "831611555a330d1412fc5497d46f47073c8633eb40f0513b842549b680771bea" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "82cfae761e5c0f68551f01def9e993c41545f7fc8d731b60e1305811d7f63067" }, "full_metric_record_sha256": "8f5367a9b6e74dc5c58ce411018f61cf4869219b9035cd277cfacd500a5af958" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1381326312725095, "detail": { "status": "ok", "value": 3.1381326312725095, "unit": "cent", "relative_error": { "count": 312, "mean": -1.6578677860476494, "mae": 3.1381326312725095, "rmse": 5.2638891349738905, "abs_p95": 11.825837544617306, "abs_max": 36.7138268607604 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 10, "false_voiced_frames": 24, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.031055900621118012, "missing_voiced_frames": 10, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.451136827468872, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "3ab09914cd862b8bc66ba18d8d21ecf9a03b6609fa2dd7e5a50e5a839629aea3" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "546be5acbc36a76c03db15837211519607b7c74777715eba9bb6794184b556ca" }, "full_metric_record_sha256": "74c8d61c23e0eba91f523a95ea21e8dcc0d0276b1efea3c7262e6074dd91a31d" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.1321596786652695, "detail": { "status": "ok", "value": 3.1321596786652695, "unit": "cent", "relative_error": { "count": 311, "mean": -1.658628164560818, "mae": 3.1321596786652695, "rmse": 5.2736616957753455, "abs_p95": 11.184868048510168, "abs_max": 38.536831153381264 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 11, "false_voiced_frames": 26, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.034161490683229816, "missing_voiced_frames": 11, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.437601327896118, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "6786194c0c460344d01a8d12083542ddfbf10dff2c22b9c292cc9df018abf47d" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b45a6380b5c8f17818af0dfae734ac4982a0ee6a4490f8dfa38759fe2f4e2cb5" }, "full_metric_record_sha256": "6eb6289382030ce00ad6beb16a174f653182d911929267dbda8324525ab3a45c" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.0796071643299365, "detail": { "status": "ok", "value": 3.0796071643299365, "unit": "cent", "relative_error": { "count": 312, "mean": -1.361792135535798, "mae": 3.0796071643299365, "rmse": 4.950780142403197, "abs_p95": 11.189608485615803, "abs_max": 28.921304375299076 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 323, "expected_unvoiced_frames": 270, "missing_voiced_frames": 11, "false_voiced_frames": 25, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.034055727554179564, "missing_voiced_frames": 11, "expected_voiced_frames": 323, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.3102355003356934, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "b96f0846f2b56b8f9785769ba8039796aa91ab7f084477d3f8fb99bcef74217d" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "dd0c86f3d424bc48154feff1e81ed6de6d4108469e419c8fcb13f5642277f77c" }, "full_metric_record_sha256": "0011936c20ba257cbd83027c7575bbf55846a192c153ff73ef72dd4bcfd81e71" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.057533845695262, "detail": { "status": "ok", "value": 3.057533845695262, "unit": "cent", "relative_error": { "count": 311, "mean": -1.2339852478175366, "mae": 3.057533845695262, "rmse": 4.926012989436227, "abs_p95": 11.084489961607677, "abs_max": 29.87996269035123 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 321, "expected_unvoiced_frames": 272, "missing_voiced_frames": 10, "false_voiced_frames": 24, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.03115264797507788, "missing_voiced_frames": 10, "expected_voiced_frames": 321, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5721683502197266, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "beb16365d019943c464287cd31fe8c4428621dd677f3eee639f4aab8616ab695" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "e0752efd9d526392bf75afa41d5a3032ad67731c86a028a46fa73f2566f68aa2" }, "full_metric_record_sha256": "7a408cb4b8214cb8aae44418856efdc7c54cae25233e11bba91702a1be537948" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.114122960661921, "detail": { "status": "ok", "value": 3.114122960661921, "unit": "cent", "relative_error": { "count": 312, "mean": -1.335792576369537, "mae": 3.114122960661921, "rmse": 5.096649965294444, "abs_p95": 11.3653674137189, "abs_max": 28.705542613903035 }, "absolute_target_error": null, "total_frames": 593, "expected_voiced_frames": 322, "expected_unvoiced_frames": 271, "missing_voiced_frames": 10, "false_voiced_frames": 25, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.031055900621118012, "missing_voiced_frames": 10, "expected_voiced_frames": 322, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6295530796051025, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "unavailable", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": null, "reason": "official_excerpt_ground_truth_not_prepared", "source_samples": 266320, "excerpt_samples": 96000 } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "bff6e00207b265cf7d4bc6cab4d1d21c8cc9ac08e896a483e5c82368ace56637" }, "native_reference": { "bytes": 768128, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 96000, "sha256": "4b45a182c576a67cdcab118faccba5276fdf67db1ce2539efa7050809fd74bce" }, "full_metric_record_sha256": "ecee139e9b8429fb038570e85ef2e3853c2e93b1cb997a6b965d9eab18b8d0ab" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.5099809169769287, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "7f9abeb916ef05c94c580ebb0a99569a11b9aaa7818162730d50f5abb46b2dbb" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "80bf274d7883bff30b22f1eae1ea2a40cb7d3f7fdc0bc75162dc90e44faa6cc8" }, "native_reference": null, "full_metric_record_sha256": "93af44f7a57626ff76d175f707d8f8afafff620b97986756b0c2277c36aa0126" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.701678514480591, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "803e26e9574c2d691c3e309777350ce3c70609be15eb83c3148d7f185c4bfde8" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "dcd1ac6a72f272d6fc7c1fa80f103ef2d5b3ce3e2829883bc3a331f60018de8b" }, "native_reference": null, "full_metric_record_sha256": "7117ec5e8ed38b26f831b7ab41495452d4a46fa3f4f237c9b8565ea8303da8fc" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.690369129180908, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "2d2bccb5778f2d002390f02e34c1630d5513d62bb4990a0f9510a30b0805aa38" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "949f626d7910d668b396b77782f8be4f4609bfac31d94289026258b8bb674406" }, "native_reference": null, "full_metric_record_sha256": "12b5e2175ed336e36e4024cd2a7afab57b52af5d16df5856a413fee685fd76c2" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.676973581314087, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "a26013c2fde7c6c78f17fb6f62ae3e36b875bc0563936ab8b6f84c45b76f2457" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "d72575fda7a9e8db365de1c65a8e9b355296ef98e347a1c3226524f9593f4159" }, "native_reference": null, "full_metric_record_sha256": "f1a7e424123148be7529629f74c834bdb9e486d3e4fd02cb40583fd2766884c3" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6481924057006836, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "bbcecff32e910ae6db84a70fb884e2a2d6dfb6d03661a4a585b9e62a23396ee2" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4be53edf46fd57f9609b6ce92a1aab70cdacb467ad6343ff715c09f767c2a743" }, "native_reference": null, "full_metric_record_sha256": "d74fefeba5b811e55c33fb8efc2ad85dfaab2f4bb64ddf1fe7a8d601387362da" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "unvoiced_miss_rate": { "status": "not_applicable", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": null, "reason": "native_no_edit", "scope": "No self-identity calculation; native pitch harm is unmeasured" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.6758594512939453, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "9bbfb80bab4b2ae0246319c49b91215aff0a321ee8001709bd4b7bba38e43dc9" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "36fdb8b277d1d4bbbecea19a534132c8cf66a9df5d90aad10e00c30fbc1497f3" }, "native_reference": null, "full_metric_record_sha256": "f3e93f1610f7c74e4a17d1ffb9552b5b557e7255c7cb3d8c375a1c5610159bc9" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.259673326757947, "detail": { "status": "ok", "value": 2.259673326757947, "unit": "cent", "relative_error": { "count": 75, "mean": -0.5454768111343742, "mae": 2.259673326757947, "rmse": 3.5085296645449118, "abs_p95": 6.635522112281005, "abs_max": 15.181313148444394 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 102, "expected_unvoiced_frames": 300, "missing_voiced_frames": 27, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.2647058823529412, "missing_voiced_frames": 27, "expected_voiced_frames": 102, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.549450635910034, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "3cd78dd43e1558e253df9d3fcdf4bff2da80bfafef15da54ba04e7546a01c53b" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "5f8ac38b5291f8debac169ad194b32c2528998f3b21c63de76ea4fcad1970559" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "80bf274d7883bff30b22f1eae1ea2a40cb7d3f7fdc0bc75162dc90e44faa6cc8" }, "full_metric_record_sha256": "98759df4e3be417aec0bfa924e0d4d5460549575a56cce1882de0dc641a8cfd3" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2128534192856892, "detail": { "status": "ok", "value": 2.2128534192856892, "unit": "cent", "relative_error": { "count": 75, "mean": -0.44140336561609456, "mae": 2.2128534192856892, "rmse": 3.471436076960112, "abs_p95": 6.546349544123397, "abs_max": 15.163471099771304 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 101, "expected_unvoiced_frames": 301, "missing_voiced_frames": 26, "false_voiced_frames": 13, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "missing_voiced_frames": 26, "expected_voiced_frames": 101, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7369284629821777, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "fe05fba702bd2c906a408eaee734d5264cb893fac1df0d6986d57e15ec2a9929" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4996340607c4cc19fc9ac2b9c3af1076d309a22c436c13e51e71de4be9774c34" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "dcd1ac6a72f272d6fc7c1fa80f103ef2d5b3ce3e2829883bc3a331f60018de8b" }, "full_metric_record_sha256": "340d1503d6fbe6d190e18cc9aabfd986f31b0eaa12b172da5fd6dda4aeb0ad1a" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.195986434538467, "detail": { "status": "ok", "value": 2.195986434538467, "unit": "cent", "relative_error": { "count": 75, "mean": -0.45646123288225243, "mae": 2.195986434538467, "rmse": 3.4663161054231306, "abs_p95": 6.542673239883315, "abs_max": 15.164641842881029 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 101, "expected_unvoiced_frames": 301, "missing_voiced_frames": 26, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25742574257425743, "missing_voiced_frames": 26, "expected_voiced_frames": 101, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7349491119384766, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "ad3063cafd1dc583d2fd434530b604ff8a1f3b11138387b55916acffa2b40672" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4813534b07e440e59d21780658c3e525df5c694ba786c2940867140d6113e5d0" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "949f626d7910d668b396b77782f8be4f4609bfac31d94289026258b8bb674406" }, "full_metric_record_sha256": "78e56f0dda4d4ffa7565039757205dde69d1d811a37180726948d5e77003d9c7" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.46836635206918, "detail": { "status": "ok", "value": 2.46836635206918, "unit": "cent", "relative_error": { "count": 80, "mean": -0.5016878395136467, "mae": 2.46836635206918, "rmse": 4.088963980608552, "abs_p95": 7.107364470545497, "abs_max": 20.02977596675289 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 106, "expected_unvoiced_frames": 296, "missing_voiced_frames": 26, "false_voiced_frames": 7, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.24528301886792453, "missing_voiced_frames": 26, "expected_voiced_frames": 106, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.657428503036499, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "38ceefe4c47f4189cc985f168477dc4033711f722b61e1f0fb8e4e281c2e7638" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "b1c6a99b922e64084c3bcbd428cc3fce45a43dd63410c1ccc7b2189d87daa79c" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "d72575fda7a9e8db365de1c65a8e9b355296ef98e347a1c3226524f9593f4159" }, "full_metric_record_sha256": "8e340e8a55cebb7813b7d188eceb38304ac2e0053578ead141c241bc41c28a55" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2404758132579286, "detail": { "status": "ok", "value": 2.2404758132579286, "unit": "cent", "relative_error": { "count": 78, "mean": -0.43060857075105585, "mae": 2.2404758132579286, "rmse": 3.4418933617562013, "abs_p95": 6.43206791341221, "abs_max": 15.358338797611173 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 104, "expected_unvoiced_frames": 298, "missing_voiced_frames": 26, "false_voiced_frames": 8, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25, "missing_voiced_frames": 26, "expected_voiced_frames": 104, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.753650426864624, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "bb9cad0324721013c70a65c866740418378d7734f333c0e7443dccf6fd62f27f" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "ec944cdc276ea33c8b75991ecfa6164d17c52ca1ccf47b9ddae76f197078e162" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4be53edf46fd57f9609b6ce92a1aab70cdacb467ad6343ff715c09f767c2a743" }, "full_metric_record_sha256": "ae4dc783da1a483b5a026f45da028d26a798bb44b20f81075cf62e5151018a4a" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 2.2368476702651754, "detail": { "status": "ok", "value": 2.2368476702651754, "unit": "cent", "relative_error": { "count": 78, "mean": -0.4345879213360591, "mae": 2.2368476702651754, "rmse": 3.4386318476051443, "abs_p95": 6.428174169995155, "abs_max": 15.363798005213118 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 104, "expected_unvoiced_frames": 298, "missing_voiced_frames": 26, "false_voiced_frames": 8, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.25, "missing_voiced_frames": 26, "expected_voiced_frames": 104, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.7262051105499268, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "04cc902d1b620846eef81ae13501a263094e9ef52f094bc0f132a85e5bac7767" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "3ab764980d96f338357e648a94fd46a5c0d8fb8e3b014786b4dbd5b4503240a5" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "36fdb8b277d1d4bbbecea19a534132c8cf66a9df5d90aad10e00c30fbc1497f3" }, "full_metric_record_sha256": "ec9a8cbb534f9fa59a7b588d534960a19f8c0091329121e92687980e0cc55c10" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.493319330572254, "detail": { "status": "ok", "value": 3.493319330572254, "unit": "cent", "relative_error": { "count": 100, "mean": -0.4423237162103609, "mae": 3.493319330572254, "rmse": 4.35530860986089, "abs_p95": 8.473038186259009, "abs_max": 14.909749882988155 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 102, "expected_unvoiced_frames": 300, "missing_voiced_frames": 2, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.0196078431372549, "missing_voiced_frames": 2, "expected_voiced_frames": 102, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.125915288925171, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "68b1deddc78d8cd6bb6d636bd74f30a7ede1d8ccf940d40ce5d5bc1a70ea20f7" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "853d3bf3d3d2c31934f366de3a4411d89890cbb2bdf551b5e388dd024fc612ef" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "80bf274d7883bff30b22f1eae1ea2a40cb7d3f7fdc0bc75162dc90e44faa6cc8" }, "full_metric_record_sha256": "6a72cd685dbeb7606aa16a5048a6647dabfc24ae49a24f5712bc680de9f74718" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.480406529246455, "detail": { "status": "ok", "value": 3.480406529246455, "unit": "cent", "relative_error": { "count": 100, "mean": -0.3440104263145674, "mae": 3.480406529246455, "rmse": 4.381175428705578, "abs_p95": 8.64079519198032, "abs_max": 14.151138048696556 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 101, "expected_unvoiced_frames": 301, "missing_voiced_frames": 1, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "missing_voiced_frames": 1, "expected_voiced_frames": 101, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.210275650024414, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "4b2be0a204af4f1e64ff6aecb97a8d82518caca375215925765502fcdf5f5909" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "67a0bf877283e452676529b80fd0e8169f52c5ddc9d3e95a8fc3e3a309fa7276" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "dcd1ac6a72f272d6fc7c1fa80f103ef2d5b3ce3e2829883bc3a331f60018de8b" }, "full_metric_record_sha256": "fc2f8a20366d822927eb47d63083d621c84902e5efbab8816e9213acbf732b87" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.4779592501715384, "detail": { "status": "ok", "value": 3.4779592501715384, "unit": "cent", "relative_error": { "count": 100, "mean": -0.33842833444227155, "mae": 3.4779592501715384, "rmse": 4.376063942711822, "abs_p95": 8.707415465525212, "abs_max": 14.038311265186683 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 101, "expected_unvoiced_frames": 301, "missing_voiced_frames": 1, "false_voiced_frames": 12, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.009900990099009901, "missing_voiced_frames": 1, "expected_voiced_frames": 101, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.258554458618164, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "5e24810d00fa04ca950b10fd48cc9ecd14e5ab02887bfd5f82219fd37cc6ccfa" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "a21ddb6e0f662806c84fd2d89b10a14e1cc382f76091a49a9470c2a3a7a88e3f" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "949f626d7910d668b396b77782f8be4f4609bfac31d94289026258b8bb674406" }, "full_metric_record_sha256": "2f1e6c9ca49e5cc7867332e683295ee52681e3647e6d0915cd6d20b554794630" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "PML", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.440849671679332, "detail": { "status": "ok", "value": 3.440849671679332, "unit": "cent", "relative_error": { "count": 97, "mean": -0.03824528940603345, "mae": 3.440849671679332, "rmse": 4.399846731835141, "abs_p95": 8.345020415272172, "abs_max": 12.7063166193718 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 106, "expected_unvoiced_frames": 296, "missing_voiced_frames": 9, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.08490566037735849, "missing_voiced_frames": 9, "expected_voiced_frames": 106, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2640607357025146, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "f250bffed073af7a4c4c0e1f4f8fe34dcea0664075be5ccc7b28bea3130ed8b8" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "53ef5dc779c5e814bb541bdd59ea0c0fb48f663ebd0e3b0af516119e62eff5e9" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "d72575fda7a9e8db365de1c65a8e9b355296ef98e347a1c3226524f9593f4159" }, "full_metric_record_sha256": "18bcb76a4c0e7294f4f3e16966a15299c7b011d1e183c0ac3994fa49969a434f" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.5121496057626467, "detail": { "status": "ok", "value": 3.5121496057626467, "unit": "cent", "relative_error": { "count": 99, "mean": 0.04667190094643607, "mae": 3.5121496057626467, "rmse": 4.424143652473062, "abs_p95": 8.490553475650236, "abs_max": 12.948111521596946 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 104, "expected_unvoiced_frames": 298, "missing_voiced_frames": 5, "false_voiced_frames": 11, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04807692307692308, "missing_voiced_frames": 5, "expected_voiced_frames": 104, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.0977516174316406, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "814bd9f4471cccc53ea8601b3b7f341d7c0579d53a222390d24e29245ace7048" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "66464f8e5e7f66373ebae514cb028af4a75cf8897a3ebf87b1857623e6e1a250" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4be53edf46fd57f9609b6ce92a1aab70cdacb467ad6343ff715c09f767c2a743" }, "full_metric_record_sha256": "c7b3b050a9ed6c9dd50b2bd0528ec540900ca49e8c73baba427603e7f2270ff2" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "status": "recorded", "quality": { "relative_f0_mae_cents": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 3.565902525944446, "detail": { "status": "ok", "value": 3.565902525944446, "unit": "cent", "relative_error": { "count": 99, "mean": 0.07716565402395129, "mae": 3.565902525944446, "rmse": 4.524563121905368, "abs_p95": 8.545090754216673, "abs_max": 12.713155382879798 }, "absolute_target_error": null, "total_frames": 402, "expected_voiced_frames": 104, "expected_unvoiced_frames": 298, "missing_voiced_frames": 5, "false_voiced_frames": 10, "unmapped_frames": 0, "out_of_range_frames": 0, "gross_error_over_600_cents": 0, "octave_error_frames": 0, "estimator": "independent-yin-difference-v2-prefiltered", "limits": "65〜700 Hz、80 ms窓・10 ms間隔。自然音声の絶対F0真値ではない" } }, "unvoiced_miss_rate": { "status": "available", "method": "existing-YIN65-700Hz-80ms-10ms-same-variant-native-reference-v1", "value": 0.04807692307692308, "missing_voiced_frames": 5, "expected_voiced_frames": 104, "definition": "missing_voiced_frames / expected_voiced_frames; fraction, not percent" }, "automatic_mos": { "status": "available", "method": "UTMOSv2-fusion_stage3-fold0-seed20261012", "value": 3.2079708576202393, "detail": { "reference_used": false, "extraction_seed_reset_each_call": true } }, "recognition": { "status": "available", "method": "wav2vec2-base-960h-22aad52d435eb6dbaf354bdad9b0da84ce7d6156-incremental-CER-WER", "value": 0.0, "detail": { "status": "ok", "value": 0.0, "unit": "CER差", "scores": { "cer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 40, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 }, "wer": { "status": "ok", "reference_errors": 0, "output_errors": 0, "truth_tokens": 11, "reference_rate": 0.0, "output_rate": 0.0, "delta": 0.0 } }, "reference_usable": true, "maximum_reference_cer": 0.2, "normalization": "nfkc_casefold_whitespace", "recognizer": "facebook/wav2vec2-base-960h", "evidence_kind": "asr_output" }, "sidecar_sha256": "f57553f563e9fab06427955da2dbd664d372d2b892a08e33abecc84aae931a16" } }, "normalization": false, "clipping": false, "alignment": false, "silence_removal": false, "source_periodic_AM_separation": "not_attempted", "waveform": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "4355e40ee25065c7b2f2fc11f03578ec8b22387c92e76ed345121d2ad778c3e0" }, "native_reference": { "bytes": 523648, "dtype": "float64", "normalization_or_clipping": "none", "sample_rate": 16000, "samples": 65440, "sha256": "36fdb8b277d1d4bbbecea19a534132c8cf66a9df5d90aad10e00c30fbc1497f3" }, "full_metric_record_sha256": "19c7cb54e2ff9be3abdf53e7afdd727d5e270227a8282007c2c71a34787deb9a" } ], "comparisons": [ { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.024420738220214844, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.057253360748291016, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.20004677772521973, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.2585625648498535, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.012205839157104492, "relative_f0_mae_cents": -0.024325713096799095, "unvoiced_miss_rate": 1.0947204968944106 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.08658981323242188, "relative_f0_mae_cents": -0.028734622941615218, "unvoiced_miss_rate": 1.5586419753086445 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.08093690872192383, "relative_f0_mae_cents": 0.00452711194506783, "unvoiced_miss_rate": -0.13882083945778168 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.014464139938354492, "relative_f0_mae_cents": 0.004267486535739273, "unvoiced_miss_rate": -0.13882083945778168 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.0307619571685791, "relative_f0_mae_cents": -0.07518098879540425, "unvoiced_miss_rate": 0.5822981366459632 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.19028019905090332, "relative_f0_mae_cents": -0.04374077498280737, "unvoiced_miss_rate": 0.5401234567901231 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1010887622833252, "relative_f0_mae_cents": 0.014539253869237712, "unvoiced_miss_rate": -0.03674669279764842 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.19679999351501465, "relative_f0_mae_cents": -0.00876325548586232, "unvoiced_miss_rate": -0.6736893679568839 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.07898068428039551, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.08323788642883301, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.0922088623046875, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.17528128623962402, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.11393952369689941, "relative_f0_mae_cents": -0.04673213932794562, "unvoiced_miss_rate": 0.0 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1572864055633545, "relative_f0_mae_cents": -0.042933598103344295, "unvoiced_miss_rate": 0.0 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.23297953605651855, "relative_f0_mae_cents": 0.01501079817986506, "unvoiced_miss_rate": -0.7516293230578952 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.2781510353088379, "relative_f0_mae_cents": 0.023728374092171656, "unvoiced_miss_rate": -0.7516293230578952 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.03378486633300781, "relative_f0_mae_cents": -0.11884320476810606, "unvoiced_miss_rate": -0.7092198581560281 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.13746428489685059, "relative_f0_mae_cents": -0.10699433996686203, "unvoiced_miss_rate": -0.7092198581560281 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.03599405288696289, "relative_f0_mae_cents": 0.10160740234052623, "unvoiced_miss_rate": -1.869559012416154 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.002937793731689453, "relative_f0_mae_cents": 0.2805916569023075, "unvoiced_miss_rate": -3.2681604110175533 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.06608939170837402, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.09217023849487305, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.15592575073242188, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": true, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.12124156951904297, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.09690690040588379, "relative_f0_mae_cents": -0.011997501423662627, "unvoiced_miss_rate": 0.8547008547008544 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.14860272407531738, "relative_f0_mae_cents": -0.025959485394397408, "unvoiced_miss_rate": 0.8547008547008544 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.18472695350646973, "relative_f0_mae_cents": 4.888692673909232, "unvoiced_miss_rate": 0.5276848755109614 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.19600343704223633, "relative_f0_mae_cents": 4.556269930856767, "unvoiced_miss_rate": -0.327015979189893 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.05991172790527344, "relative_f0_mae_cents": 0.836259907777555, "unvoiced_miss_rate": -0.8547008547008544 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.016646146774291992, "relative_f0_mae_cents": 0.6801995600491546, "unvoiced_miss_rate": 0.0 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.01998758316040039, "relative_f0_mae_cents": 4.6439318897757325, "unvoiced_miss_rate": 0.7209215904868074 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.01812291145324707, "relative_f0_mae_cents": 4.229995263283303, "unvoiced_miss_rate": 1.575622445187662 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.037071943283081055, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1291825771331787, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.022644519805908203, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.05150485038757324, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.029789447784423828, "relative_f0_mae_cents": 0.09445744831170755, "unvoiced_miss_rate": -0.5807357267849935 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.05213737487792969, "relative_f0_mae_cents": 0.05696436118087611, "unvoiced_miss_rate": -0.27017672057381426 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.014386892318725586, "relative_f0_mae_cents": -0.03548351558617302, "unvoiced_miss_rate": -0.8554922214827885 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.06804609298706055, "relative_f0_mae_cents": -0.04292259431989809, "unvoiced_miss_rate": -0.27113820356517854 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.118927001953125, "relative_f0_mae_cents": 0.029408054322312527, "unvoiced_miss_rate": -0.6095802165259696 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1053915023803711, "relative_f0_mae_cents": 0.02343510171507246, "unvoiced_miss_rate": -0.29902121031478923 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.2619328498840332, "relative_f0_mae_cents": -0.022073318634674433, "unvoiced_miss_rate": -0.2903079579101683 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.3193175792694092, "relative_f0_mae_cents": 0.03451579633198465, "unvoiced_miss_rate": -0.29998269330615523 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": true, "recognition_available": false, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1916975975036621, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.1803882122039795, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.02878117561340332, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.0011141300201416016, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": null, "unvoiced_miss_rate": null }, "harm_observed": false, "required_observations_missing": true, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.18747782707214355, "relative_f0_mae_cents": -0.04681990747225795, "unvoiced_miss_rate": -0.7280139778683747 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.18549847602844238, "relative_f0_mae_cents": -0.0636868922194802, "unvoiced_miss_rate": -0.7280139778683747 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.096221923828125, "relative_f0_mae_cents": -0.22789053881125154, "unvoiced_miss_rate": 0.47169811320754707 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.06877660751342773, "relative_f0_mae_cents": -0.2315186818040047, "unvoiced_miss_rate": 0.47169811320754707 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.08436036109924316, "relative_f0_mae_cents": -0.012912801325799084, "unvoiced_miss_rate": -0.9706853038245 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": 0.13263916969299316, "relative_f0_mae_cents": -0.015360080400715681, "unvoiced_miss_rate": -0.9706853038245 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.16630911827087402, "relative_f0_mae_cents": 0.07129993408331448, "unvoiced_miss_rate": -3.6828737300435406 }, "harm_flags": { "automatic_mos": true, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": true, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "pair_status": "recorded", "deltas_to_same_pitch_seed_PML": { "automatic_mos": -0.05608987808227539, "relative_f0_mae_cents": 0.12505285426511392, "unvoiced_miss_rate": -3.6828737300435406 }, "harm_flags": { "automatic_mos": false, "relative_f0_mae_cents": false, "unvoiced_miss_rate": false }, "harm_observed": false, "required_observations_missing": false, "recognition_available": true, "not_an_acoustic_or_overall_classification": true } ], "combined_saved_acoustic_and_additional_metric_status": { "aggregation": "Boolean OR of the unchanged acoustic warning and the added metric warning; missingness and unresolved source-AM risk remain separate. This is not a perceptual-quality or adoption classification.", "counts": { "RAW_FULL": { "conditions": 30, "saved_acoustic_harm_present": 25, "additional_metric_harm_present": 0, "any_observed_harm_present": 25, "additional_required_observations_missing": 22 }, "RAW_SHAPE": { "conditions": 30, "saved_acoustic_harm_present": 20, "additional_metric_harm_present": 2, "any_observed_harm_present": 20, "additional_required_observations_missing": 22 } }, "conditions": [ { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0008", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1188-133604-0003", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": true, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_121-121726-0012", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 0, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": true, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": -3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": false, "additional_metric_harm_present": false, "any_observed_harm_present": false, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 0, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_SHAPE", "saved_acoustic_harm_present": true, "additional_metric_harm_present": true, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" }, { "case": "librispeech_1089-134686-0032", "pitch_semitones": 3, "seed": 20261015, "variant": "RAW_FULL", "saved_acoustic_harm_present": true, "additional_metric_harm_present": false, "any_observed_harm_present": true, "additional_required_observations_missing": false, "source_periodic_AM_separation": "not_attempted" } ], "original_relative_crossing_supplement": { "scope": "Post hoc description of the saved RAW_FULL PML-relative crossing warning entries. No thresholds or flags are changed.", "PML_relative_warning_band_entries": 38, "among_these_original_relative_absolute_difference_at_most_1ms": 36, "among_these_original_relative_absolute_difference_over_1ms": [ { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "band_Hz": [ 1500, 3000 ], "output_minus_original_ms": 6.999999999999673 }, { "case": "librispeech_1221-135766-0005", "pitch_semitones": 0, "seed": 0, "band_Hz": [ 3000, 8000 ], "output_minus_original_ms": 4.0000000000000036 } ], "comparison_note": "The numerical tolerance above only handles binary floating-point representation of a saved 1 ms grid difference.", "strong_power_drop_distinct_conditions": 8 } }, "model_provenance": { "utmos": { "algorithm": "UTMOSv2-fusion_stage3-fold0", "source_commit": "cc2700db57bb83ee13dc31ebe1b868c254e15d09", "predict_dataset": "bvcc", "extraction_seed": 20261012, "extraction_repetitions": 1, "device": "cpu", "remove_silent_section": false, "calibration": "対象ボコーダーでの人間評定校正は未実施", "assets": { "checkpoint": { "bytes": 818531314, "sha256": "c8149d988e4bbf3f347e6966b5d769de347a5f8c59ffca1dc4bd4bf5b8585e57", "expected_sha256": "c8149d988e4bbf3f347e6966b5d769de347a5f8c59ffca1dc4bd4bf5b8585e57" }, "config": { "bytes": 1842, "sha256": "4937977e24d12d1bba70cdce8709c3c04807a8e4ae8ddac4229c48c436ae99ae", "expected_sha256": "4937977e24d12d1bba70cdce8709c3c04807a8e4ae8ddac4229c48c436ae99ae" }, "preprocessor_config": { "bytes": 159, "sha256": "b225d617c025463b9e157e06afea8b90dc7078fc70b013c533328423e0486b4a", "expected_sha256": "b225d617c025463b9e157e06afea8b90dc7078fc70b013c533328423e0486b4a" } }, "packages": { "utmosv2": "1.3.1.dev0", "torch": "2.8.0+cpu", "transformers": "4.51.3", "timm": "1.0.15", "librosa": "0.11.0" } }, "recognition": { "recognizer": "facebook/wav2vec2-base-960h", "revision": "22aad52d435eb6dbaf354bdad9b0da84ce7d6156", "decoding": "argmax logits then CTC collapse; no language model; batch_size=1", "processor_internal_normalization": true, "external_normalization": false, "model_card_url": "https://huggingface.co/facebook/wav2vec2-base-960h/blob/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/README.md", "asset_sha256": { "config.json": "d3ec255c063d9f95057b553b19c20135b259875834a4fe9deb218a6be25b4cf3", "preprocessor_config.json": "b225d617c025463b9e157e06afea8b90dc7078fc70b013c533328423e0486b4a", "tokenizer_config.json": "dc790594f5bc351a4311c6624f40acd95850d4aaf2a5cb3c656c9b610720b608", "special_tokens_map.json": "bb7068de1150661a10b55f9e4b12a0e77af8bf91f5e45e1b58afaf1d0e17f675", "vocab.json": "19727f8944fe6459fc3f240ae2c198395b740f6a029bd23e06656266b83bcf64", "README.md": "fc52a0adfa64e446800f9944c6cf9b90da3127aba548860ea5714516ec8c18c7", "model.safetensors": "8aa76ab2243c81747a1f832954586bc566090c83a0ac167df6f31f0fa917d74a" }, "asset_bytes": { "config.json": 1596, "preprocessor_config.json": 159, "tokenizer_config.json": 163, "special_tokens_map.json": 85, "vocab.json": 291, "README.md": 4431, "model.safetensors": 377607901 }, "asset_urls": { "config.json": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/config.json", "preprocessor_config.json": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/preprocessor_config.json", "tokenizer_config.json": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/tokenizer_config.json", "special_tokens_map.json": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/special_tokens_map.json", "vocab.json": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/vocab.json", "README.md": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/README.md", "model.safetensors": "https://huggingface.co/facebook/wav2vec2-base-960h/resolve/22aad52d435eb6dbaf354bdad9b0da84ce7d6156/model.safetensors" }, "source_sha256": { "scripts/added_metrics_asr.py": "7ef7f8b8ff3bd3eb3d96967afad0b87e219ebeb023714f34fb786ed2a5707db1" }, "assessment_source_sha256": { "__init__.py": "37126abeb4fe6f6ce927d264c42a1b2ab934eb9b11a25faa8165d1685c37f63c", "__main__.py": "ce535ce8695e480492699a7a859ad36daea0e8cf0f192f8dbbd76ed00891a2e0", "acoustics.py": "50737f8804f433b09737442e30c224a00e5f30f12583d38c4b7465b40d9669a3", "common.py": "ab938071273d5d475df871e521cba4ebb21e6d352c9e52d994c04cf71a957829", "controls.py": "8d632e7e95f5d674f034e865a4f532247cec41c548e27cc54dd75d09504653f3", "curves.py": "4a794d7d7a94666f298b46ff3c6b6ac943776660ad2ea4f1174fc922140006f1", "probes.py": "b1bced34227ce72b5e5fe640cf39c761172aeaa0ef9341829f19bdaf4a984f39", "representation.py": "84ea8e9de44e797c99066e36ce862a38759a96a4c8e0671f57c5a70c8c0b12b6", "runner.py": "44a522b612cbaae6f0dad91153cd315bc441964feb3b6faab24938a95fb3e677" } } }, "evaluator_source_provenance": { "fixed_metric_source_freeze_sha256": "62b8d6680fa61c29ce983b3a27cdb00696010e04276b16e14cc58f96c2476baf", "fixed_metric_source_file_sha256": { "integrity.py": "489d3aaab24bc758f52449b57f91265b744ec74af3150b3e9facb86151263c19", "metric_backends.py": "5556c861b6142c6bcae95107e5bd9840ec4c21bc831fa42bb768f2e935e11925", "metric_contract.py": "bbc1ce78ec0455c600e65606363ece15e35e6e8cd94f7e208b1711139d13de6d", "plan.json": "ed4e2707db1653389d5d13dcd0b29341b6acf95d03c0c9a965c2a328bf03737c", "run_saved_metrics.py": "bd01f4e5b33638429595a4581ac08850938d839ba60767864036f5f85779ee82" }, "frozen_runtime_source_manifest_sha256": "8528307f71e77cbe27861f5457c36622216ad0f20f60d3d6529f820c5d0094cf", "frozen_evaluator_repository_file_sha256": [ { "artifact_name": "BASELINE.json", "sha256": "02fa0c4315f4401a817ac38d0c7ed523eb5e2ef43523485596d52a72bbdf3a92" }, { "artifact_name": "pyproject.toml", "sha256": "f6ab6f877acd61ef43b642081d2441b8b61d80c2da599f5cffa910797884252d" }, { "artifact_name": "scripts/added_metrics_asr.py", "sha256": "7ef7f8b8ff3bd3eb3d96967afad0b87e219ebeb023714f34fb786ed2a5707db1" }, { "artifact_name": "scripts/added_metrics_asr_summary.py", "sha256": "80d5182d06cf3f9aca3a19ae9489ecd02d1a551d63d56255ec16fc0217845cb6" }, { "artifact_name": "scripts/added_metrics_assessment.py", "sha256": "9b9a008e0019f280a5b17fc8f81de0eb2903f87f522a4d0247808746ed608c80" }, { "artifact_name": "scripts/added_metrics_assessment_run.py", "sha256": "af6e37f26477ac038ae5b0a45c34706907967fd07d1b820aba765195e2a10331" }, { "artifact_name": "scripts/added_metrics_assessment_summary.py", "sha256": "72cee89719996a1fd8f2436d11ac84b8171518835d9dc8724ea81654c2b75641" }, { "artifact_name": "scripts/added_metrics_quality.py", "sha256": "8087feec9501fc6ca8d953768a8bc1733c28dabc32b8fc719ad9599f4c3ef7cb" }, { "artifact_name": "scripts/added_metrics_report.py", "sha256": "4cfca82585aa4433edb68e9e62ebbb45d6945c34281f05e5bfa966aee39f5979" }, { "artifact_name": "scripts/added_metrics_report_data.py", "sha256": "3f034f800ebdcad252d8dcb9329b9e42a67f8e76a70a8b37ce219866320c3f6e" }, { "artifact_name": "scripts/added_metrics_report_html.py", "sha256": "7149edd50b78f4e4ef0bbaebf9c9888f94838a811e5274fef9057a6e3f62f015" }, { "artifact_name": "scripts/added_metrics_report_sources.py", "sha256": "11d5e3d944be1621060e305a2dfd1c8d2b1d35c1aa5591512b69cad53edbacd2" }, { "artifact_name": "scripts/added_metrics_synthesize.py", "sha256": "9c6a31e098961d8acafe6dc208bd68122df5204881457a65898d93751c254719" }, { "artifact_name": "scripts/all_metrics_acoustic.py", "sha256": "4504db5ae90cc02f76bab8f6302b74d5d96fa0b402446609854bec195fc4e474" }, { "artifact_name": "scripts/all_metrics_asr.py", "sha256": "b6521c06a7ea28d92ef4ec58fce0c92eb0cb4c10fc835c0739303faea56ba6ab" }, { "artifact_name": "scripts/all_metrics_audio.py", "sha256": "cdcab0b6e740d5525d0c89c6e5d41b8fe6cc643a1dd6fb350a2affcc52a440dc" }, { "artifact_name": "scripts/all_metrics_audio_policy.py", "sha256": "890f6cc558604f7eeef82a888852ff487d9057dfd9c3957265628d112f2c0fdc" }, { "artifact_name": "scripts/all_metrics_base_controls.py", "sha256": "fcc28f91eaee12d9d81f70e2d17667ce7f4cf2ace8329da8db987e784281cbb7" }, { "artifact_name": "scripts/all_metrics_checkpoint.py", "sha256": "f235d550ab393134d5f79b61cdbc0e13fbedeb168dff49fa6fea52abab457b5c" }, { "artifact_name": "scripts/all_metrics_common.py", "sha256": "d9d2ea1fc7bb93a627d14db1eddda203d91cb9c0f1708d31db03b148ad67168a" }, { "artifact_name": "scripts/all_metrics_control_adapters.py", "sha256": "e0683815d17fce0a2b723732d5d2aa1eb852f81c44fe2e3eadcbc26daf1a2327" }, { "artifact_name": "scripts/all_metrics_control_eval.py", "sha256": "099814a22a7863ceb799593c55f7e5cf04efbb6f1a1df0675b36ed6b163c3bfc" }, { "artifact_name": "scripts/all_metrics_control_eval_curves.py", "sha256": "e120bb5084cf71da167cd068c8771afd2398d13893518f8c1eb4aff78926f1f1" }, { "artifact_name": "scripts/all_metrics_control_eval_equivalence.py", "sha256": "bb464d5d10b6ef1e7c28569a9f8d8a378d45582a4ca33b48c6118365efc93319" }, { "artifact_name": "scripts/all_metrics_control_eval_plan.py", "sha256": "da64f73d3bc8aba020547a5bce8ec8ad216d0ea9d456e7c474e92ebbe386b6e4" }, { "artifact_name": "scripts/all_metrics_control_eval_resume.py", "sha256": "1c672636c63879126eb0cc56a49fce49c8084cd0658c52c9e6c7dec31c76dd61" }, { "artifact_name": "scripts/all_metrics_control_eval_run.py", "sha256": "3b92ea8ce219212402652491838e3b8000b7a967ee8719ec8fc0369eb0be7a6b" }, { "artifact_name": "scripts/all_metrics_control_eval_slope.py", "sha256": "32d97162e8b8fd4a16634c10489e1302d95717ecd1a2ed2ecbb3ab56d9f73d89" }, { "artifact_name": "scripts/all_metrics_control_eval_speaker.py", "sha256": "ee0c81c65566f9dd6bb3f0c431edb0717768b1e16c48cd2b5a8837c04560ebce" }, { "artifact_name": "scripts/all_metrics_control_eval_summary.py", "sha256": "09a52aca15f6b7bd1791a86bf193339ce2ad355fd85e6ce8befbb020a3333d61" }, { "artifact_name": "scripts/all_metrics_control_eval_time.py", "sha256": "1b6ecc78d691f2cbc0405b471b410b3b6befcf800a35e21f1c82971f9b8cd86f" }, { "artifact_name": "scripts/all_metrics_controls.py", "sha256": "b8a10df3a157ea5f08ddeb4b4c8b5ddb2c4f65ee933a01686ab5e907ad244729" }, { "artifact_name": "scripts/all_metrics_datasets_abx.py", "sha256": "1fecd0f7ee92f3f4397a0edd6ef0f6f07f3dadd4a79c65b695568bc549bda0e3" }, { "artifact_name": "scripts/all_metrics_datasets_abx_contract.py", "sha256": "aba26344cd0ee2a6064c170f3c180498b81e4456a7ec095bba4c41278d6bd85e" }, { "artifact_name": "scripts/all_metrics_datasets_abx_sources.py", "sha256": "f18f06a5389622b7d444368a22d8491011faa0f421987afc9be5ed3e4f222626" }, { "artifact_name": "scripts/all_metrics_datasets_contract.py", "sha256": "4d995c8f706b51fd585b606c93e29f68c8c57c92aa4fa83c23cb375a0633a3d2" }, { "artifact_name": "scripts/all_metrics_datasets_human.py", "sha256": "6e29ddd362650607e2d1898b59717dd8a6e05cbe084c4a3c8079e383e4befef7" }, { "artifact_name": "scripts/all_metrics_datasets_jvs.py", "sha256": "24f1054752fe49f334cfca078be8142fd13264293008eaf8c94ce7131cbec6f1" }, { "artifact_name": "scripts/all_metrics_datasets_remote.py", "sha256": "03d6b0e9249e8511cc2a7990a3fa2b7041050b6ed3c56a5b520c0065cc0566e6" }, { "artifact_name": "scripts/all_metrics_datasets_train.py", "sha256": "cfbe7ee6f9da6f307d5f73f21771413903b859f806f690629aea89aa2f7b2564" }, { "artifact_name": "scripts/all_metrics_datasets_vctk.py", "sha256": "7b296ba8a4d5c6bcdc027dc44e49a4970b0b0be9242fe5f614e9ab71bdf8f506" }, { "artifact_name": "scripts/all_metrics_derive.py", "sha256": "6ac791c01dbe6bde9d1892f50f43796bac22904ce1c117a7989f164acd230f55" }, { "artifact_name": "scripts/all_metrics_feature_locked_run.py", "sha256": "c00ef0128e474def36a86a5f7bee72a0d61d3ad777eaf04841f018d59bbfcca9" }, { "artifact_name": "scripts/all_metrics_features_abx.py", "sha256": "854766030318e4325332039840dc82676c66e236e1261354923aafa6af92b6dd" }, { "artifact_name": "scripts/all_metrics_features_abx_results.py", "sha256": "1f527abea484e7530ddb3deb80f5e9b59d836cebe893a4d98211d90bb672e76d" }, { "artifact_name": "scripts/all_metrics_features_abx_statistics.py", "sha256": "2964800a12875e89204500023b35225c010c8bb16a3f3938cd1e2b5c9198359d" }, { "artifact_name": "scripts/all_metrics_features_common.py", "sha256": "e6b3a6783ef2f2085dd694ffee30af06dac0b7705e491e2c99cc02f155d8b89f" }, { "artifact_name": "scripts/all_metrics_features_readout.py", "sha256": "e20a2c269296f1aa3986232f02b449868741d2ba4e8113bad23d48b302159adf" }, { "artifact_name": "scripts/all_metrics_features_receipts.py", "sha256": "6bacc0afd2fe292a9e96cbc8301e3cb3184581a3c441f8aa6e93630050770dda" }, { "artifact_name": "scripts/all_metrics_features_run.py", "sha256": "7087d9de6aa8c5b7aeb35b069f792cd9587ff3c0020faf331e81692851b34096" }, { "artifact_name": "scripts/all_metrics_features_support.py", "sha256": "cf924c6d5dc26196a43b5d23ca58bb68628658f8f3b08d57989855bf44b6f2c0" }, { "artifact_name": "scripts/all_metrics_features_train.py", "sha256": "2bb864a20d2bc188822fd29fc6d986eed64ded9a834b2faba894f82a0cbc065e" }, { "artifact_name": "scripts/all_metrics_final_gate.py", "sha256": "1437262a4cb3aa8f77d701436959d96f67d0b278f9a058f5d82cfa47baf3d43b" }, { "artifact_name": "scripts/all_metrics_golf_control.py", "sha256": "883ff581ab8132a7c8d312c894a26eb29990cf9a71846223c2aa8d6434ff3b96" }, { "artifact_name": "scripts/all_metrics_human_data.py", "sha256": "413edc016c33b9101d22a4b46a3e3ebc3a2276b76bb0528a396cda63b4339d05" }, { "artifact_name": "scripts/all_metrics_human_features.py", "sha256": "1943b27ad7c0cb68e700e147785b4a8df117b718b33e095a6a7a5e7e9ae3427b" }, { "artifact_name": "scripts/all_metrics_human_rankings.py", "sha256": "404cade43cbec39d939b0f346fb88e03584063473c6945bea7d9176fce5c9eb5" }, { "artifact_name": "scripts/all_metrics_human_run.py", "sha256": "61fe1ac94cd4104a9c318806b59393ea5592afcd1d80a034f5ccf6318cb98532" }, { "artifact_name": "scripts/all_metrics_human_statistics.py", "sha256": "ec998755947c516e58cb1851650b7d89fcfaf79c85a9c3b8820da0f3b62bc943" }, { "artifact_name": "scripts/all_metrics_interpolation.py", "sha256": "afa53eedf0182c4611456096542b57002d1c0ff729c3f9e31b2d7e111cf89a03" }, { "artifact_name": "scripts/all_metrics_latent_evaluation.py", "sha256": "73879fbb0848182a7c79ad1f01078beb70fb2ec2c6499fec2a7d3e4491907290" }, { "artifact_name": "scripts/all_metrics_latent_evaluation_common.py", "sha256": "bc890a2cb0db62c5d3e5707da6dd230ab81d203ee1341f15b5eff708cbefce00" }, { "artifact_name": "scripts/all_metrics_latent_evaluation_resume.py", "sha256": "bf6394881eb472135af1089165ebd73fdd90276d3dd988e9feb31bb6b4e39e5f" }, { "artifact_name": "scripts/all_metrics_latent_evaluation_storage.py", "sha256": "1011459a6452ade6e29682671e1041856f6411b3ecabc6f1e439a952149deef5" }, { "artifact_name": "scripts/all_metrics_latent_evaluation_watch.py", "sha256": "b35678de0b8df7a206058d71b59e3a030584943942dcaf90074deb4a671e06e8" }, { "artifact_name": "scripts/all_metrics_latent_features.py", "sha256": "067fa971de819fc01982558117d692bb741d6d2916939406b15d9f354786eb91" }, { "artifact_name": "scripts/all_metrics_latent_groups.py", "sha256": "5e006c23d3329d75de369f9566f6b809fa3ca77057530fe8dc3642c4bd1350c8" }, { "artifact_name": "scripts/all_metrics_latent_interpolation.py", "sha256": "070200bdca42a7d92313cc9acd3447bc79cd5b06bb2143e139a97e98c42b648e" }, { "artifact_name": "scripts/all_metrics_latent_io.py", "sha256": "323b2c05836b1e44400a5e412471f61a8e18194de9c26a511b8fb06c6931fa0f" }, { "artifact_name": "scripts/all_metrics_latent_provenance.py", "sha256": "02691b5311f1ceb5bd8c7a2b0a6255dbe870860d6229259c49cd6bbe91909658" }, { "artifact_name": "scripts/all_metrics_latent_quantization.py", "sha256": "af439e7ad3581aeee044c8daa911d09be40652fe5136cac519b93904a8cda25d" }, { "artifact_name": "scripts/all_metrics_latent_queue.py", "sha256": "2c1985438d5444b6e07e7d9b45d7b0ee9496a2d6ab97098397545f0cd96d79c3" }, { "artifact_name": "scripts/all_metrics_latent_run.py", "sha256": "2ad6504ae59a13eead7de5af9eb42d574877474355586c01d286970e4f6703cb" }, { "artifact_name": "scripts/all_metrics_latent_worker.py", "sha256": "acdb1f43ac03af1617036b72a91c6bed29c97574ab00760e6e5cab75c2f41a49" }, { "artifact_name": "scripts/all_metrics_llsm_control.py", "sha256": "6ba12b6933dfda3f9d6b57f46085de652f00acd348cf1ed5bef15eb60c7ad765" }, { "artifact_name": "scripts/all_metrics_llsm_storage.py", "sha256": "c672f5f38f934ff9710a726f029a102592c3f5e9de8d25acdace76dfae64251f" }, { "artifact_name": "scripts/all_metrics_mos.py", "sha256": "fbfc7e90b1c6c133b586f4a21c25361940aeafa14c876b6891b9627069783607" }, { "artifact_name": "scripts/all_metrics_neural_controls.py", "sha256": "c6a3c150c448ec41a04befc6543bd25d4512bcc28d80cf0bf14b5d8481ac66ce" }, { "artifact_name": "scripts/all_metrics_process_receipts.py", "sha256": "84ba1ac16a32a8e7965d5d916d5b78ed1d475913138c76a6020eecd611b80dbc" }, { "artifact_name": "scripts/all_metrics_rankings.py", "sha256": "416ad40711e2680467c0de2f0fffa53806d9b857256beeea9edf186e6ba292bf" }, { "artifact_name": "scripts/all_metrics_rankings_acoustic.py", "sha256": "a6e306066c0c112ed2dfb72cbeb2c498d00b445cd8443cf645fddf3fccf3d405" }, { "artifact_name": "scripts/all_metrics_rankings_availability.py", "sha256": "e23a593a9f3e2f7d728f868a7f742bdafa7020f192824760d7aaa4cdca914822" }, { "artifact_name": "scripts/all_metrics_rankings_core.py", "sha256": "77d8eca35f00f086ca3f137181fdeb36438d6606126205ce1f12993bda844159" }, { "artifact_name": "scripts/all_metrics_rankings_coverage.py", "sha256": "5e2bf98764f53467e0f6f8f5797f7f1b5fd70b7cec26ffc58fda0a8b4ee43b66" }, { "artifact_name": "scripts/all_metrics_rankings_extra.py", "sha256": "7b208413006fa0ddb0700eb438d0171f272bd0184237256bce3f59756818de18" }, { "artifact_name": "scripts/all_metrics_rankings_html.py", "sha256": "a50e284d23dfc81990a28c6550134972cbb17c01c92ddcc09b0afe9edaa5d25e" }, { "artifact_name": "scripts/all_metrics_rankings_inputs.py", "sha256": "464ab0aed1af4de31e83db80c43857e517d9cf61667904144c28f8f08ecd42b5" }, { "artifact_name": "scripts/all_metrics_representation.py", "sha256": "973e341c131698a36fb2a2b82fd7dd4d59b97dce6c06b2f698f595771ead5da2" }, { "artifact_name": "scripts/all_metrics_resources.py", "sha256": "b142bc88c83665ebb8ad1f806569deb838ade589a0d28447ec45274008aedfbd" }, { "artifact_name": "scripts/all_metrics_run.py", "sha256": "6412256d4ba20bf575b25e750d4bd9a13dfcc6f000c6ac73fe4228d37ae3abef" }, { "artifact_name": "scripts/all_metrics_storage_policy.py", "sha256": "81f0b90ec0d0467a5fb7767df9800fc01f713d9f5e4b5559cf6db8776ccf11ac" }, { "artifact_name": "scripts/all_metrics_synthesis.py", "sha256": "8b6925078b5d05792af2593fd86dac93d726e8f9ce9fd80aa45c961683e0eb42" }, { "artifact_name": "scripts/all_metrics_vctk_inputs.py", "sha256": "ac71116457eab20b74414f4a1591078b6c6b09a9fa6fed3f6eb0393893232b58" }, { "artifact_name": "scripts/all_metrics_vctk_quality.py", "sha256": "9f687673c295e8d735dce390c2408c19bafed689c487a2583f26a4070596925e" }, { "artifact_name": "scripts/all_metrics_vctk_quality_resume.py", "sha256": "bd6855952d6cae49675282d5d1ed89c8d742762d79ff6851f06ae1e7f17366a9" }, { "artifact_name": "scripts/all_metrics_vctk_resume.py", "sha256": "e665c9225ca145bc57682b1cca007091be3497b49197954fa4e8d678e0f22cc9" }, { "artifact_name": "scripts/all_metrics_vctk_run.py", "sha256": "1099c927901d08bc3c68f4272ae0a884f49b2fac381b40edb494deed85d0e74f" }, { "artifact_name": "scripts/all_metrics_vctk_summary.py", "sha256": "03245d68aa53f3724dbef7519e25c3c179bcef3ff5c610f5ff2e5f721bbc5e6b" }, { "artifact_name": "scripts/all_metrics_vctk_synthesis.py", "sha256": "bb6977d596417bb5addb7ae660c080f4348bb5e2b942c3503a3611dcb7d8cf64" }, { "artifact_name": "scripts/all_metrics_vctk_watch.py", "sha256": "cad25b6078b729984cfb1237e4fc5a49519b4ad2816514788b8a403acac81539" }, { "artifact_name": "scripts/build_pml_package_report.py", "sha256": "513cd61ea222c70049faec1e51acc17cc7881802991e442ce949f17f63070e3c" }, { "artifact_name": "scripts/build_site.py", "sha256": "a238f1d2cd486f2c25c53d68ff0b2f637c927707b8913eb4568dd6c69b155b30" }, { "artifact_name": "scripts/check_source_size.py", "sha256": "6be81c507df2592fe1e5019171e7b5f24a06a17fe52f14c23da242bc6711216c" }, { "artifact_name": "scripts/chunk_evaluation/__init__.py", "sha256": "db8a6c2865f751d53e4275b266c8e9e9c6160cfbe65331e5b14d44960750b37d" }, { "artifact_name": "scripts/chunk_evaluation/aggregate.py", "sha256": "96307b39c07d866c385fad9b73e4d0ee1abd979bb6f05c9200b595a1722c00cd" }, { "artifact_name": "scripts/chunk_evaluation/axis_registry.py", "sha256": "3b0b4434aa9c0510dc0f13d7c07f573cf15c6afb364a3738f6b4e02b7d4ecbe4" }, { "artifact_name": "scripts/chunk_evaluation/build_axis_rows.py", "sha256": "674fe43c83ed6710e956f28a86fd58fcf50b174a079e70d3346e715cc5366d12" }, { "artifact_name": "scripts/chunk_evaluation/chunk_checkpoint.py", "sha256": "286fefe2703106e1e5ebc28a3a6c45bea3822ea65dfbd24eddd475f97c32cbcf" }, { "artifact_name": "scripts/chunk_evaluation/chunk_contracts.py", "sha256": "81dc1f5c5ef0b9939833bae56f8767e17d9d87ebf9d3e0d8d20ced8dbec5ae0d" }, { "artifact_name": "scripts/chunk_evaluation/evaluation_io.py", "sha256": "c5ed1f2ae74034dc4c1545ec874f812fcc0a8fe267714c7c0d90636dabf58ae3" }, { "artifact_name": "scripts/chunk_evaluation/freeze_chunk_plan.py", "sha256": "685c44d8fc2582153221ee8e097b7395f7c7b0cf821d09c8dd8e874ac4c31d65" }, { "artifact_name": "scripts/chunk_evaluation/freeze_chunk_waves.py", "sha256": "c73b5dac7909bd4bb24580cabc9e8a6f560df07e8938cb0534b19994eaa9b3a4" }, { "artifact_name": "scripts/chunk_evaluation/reference_context.py", "sha256": "7f2922c037090d20ce95e20e6327fee1a13a4d8e09979df4c2e5d4ea404d71fd" }, { "artifact_name": "scripts/chunk_evaluation/reference_generation.py", "sha256": "72760d26fd4e84706cd95cd69b14939301723c91acfc7699777670784243a745" }, { "artifact_name": "scripts/chunk_evaluation/report.py", "sha256": "0f1ab78dfb26a7bf644f6157588263904f01c7db9f2906c0a228b1917a585301" }, { "artifact_name": "scripts/chunk_evaluation/run_asr.py", "sha256": "02cc631a84000ec876dba3164d4f7964d995276ecc4f421789e4944d0b8ac2b1" }, { "artifact_name": "scripts/chunk_evaluation/run_metrics.py", "sha256": "cf61ea5a0e82b6994ec3ffa76b688efd2ab8a514441edb08ff73a100703f6d54" }, { "artifact_name": "scripts/embed_pml_package_report.py", "sha256": "50c43fb36d41189c6680aea3e116cfc3876df9523f32a68d3aefedfcd4aea2f5" }, { "artifact_name": "scripts/fetch_data.py", "sha256": "12d25be0313296f1edc3e71dcf57b03ddfb0cedd0c58df995c02d234254ac476" }, { "artifact_name": "scripts/noise_jvs_report.py", "sha256": "640e9621ab4941988970cc95b8408fb0f697407bba9d3eb0b6a4067e8b22d047" }, { "artifact_name": "scripts/noise_report_provenance.py", "sha256": "d9e75dbcde1e4a464de9362c1fe1d990a348521987ac35a12a12ad86a3703a76" }, { "artifact_name": "scripts/noise_screen_report.py", "sha256": "1f512461f08f40eabb0febd3a507997b100300fe4aa08babfa0e8c299a3e41c0" }, { "artifact_name": "scripts/pml_evaluation/abx/abx_inputs.py", "sha256": "30c9b20700d3dd8d1397079e98a3deed1842858b8c904f81299ac4d8c4c416a7" }, { "artifact_name": "scripts/pml_evaluation/abx/compare_abx.py", "sha256": "43dd105372e402bfcf841a4b6b991c57313bb195f3b24943ba5e73c55dc20aec" }, { "artifact_name": "scripts/pml_evaluation/abx/frozen_abx_statistics.py", "sha256": "2964800a12875e89204500023b35225c010c8bb16a3f3938cd1e2b5c9198359d" }, { "artifact_name": "scripts/pml_evaluation/abx/run_abx.py", "sha256": "2b1c81d52c15f03708a914422efa2f690ce0377235e9d26168ed92bb0db0dddf" }, { "artifact_name": "scripts/pml_evaluation/aggregate_available.py", "sha256": "5ab43a60989cc469f5e99382bd3631e92b4e63d97fee67fb5d6997f5b82e17f7" }, { "artifact_name": "scripts/pml_evaluation/aggregate_seed_variation.py", "sha256": "b0e42381615bb8aab9dcaa632ee02cb1ec9d76519cbb9bff7d848aafe4b8c8f0" }, { "artifact_name": "scripts/pml_evaluation/aggregate_timing_cohorts.py", "sha256": "e91c3e4f4209ccfbb0f688495e435ea4b2c5054cce5fe79cf11431d431fbdf2b" }, { "artifact_name": "scripts/pml_evaluation/auxiliary/aggregate_asv.py", "sha256": "04245a95d975e3881da30939dfed5e2392165a73d9f1ee5ba92a3c8101a9335d" }, { "artifact_name": "scripts/pml_evaluation/auxiliary/extract_asv.py", "sha256": "b5fd2e71f031c9feae82455e075390d9e7d41a32793a5bafb96377938cb6d09d" }, { "artifact_name": "scripts/pml_evaluation/auxiliary/human_statistics.py", "sha256": "ec998755947c516e58cb1851650b7d89fcfaf79c85a9c3b8820da0f3b62bc943" }, { "artifact_name": "scripts/pml_evaluation/extended/extract_features.py", "sha256": "e4daaed4adc097b769250ad6a1ed66eeb07dece1457737314e67bc40a9b485ef" }, { "artifact_name": "scripts/pml_evaluation/extended/measure_quantization.py", "sha256": "1242a7c9b8e1829ea63b0d59cb2c7ed9a31804caeee8d714dedaa9c35b6035c5" }, { "artifact_name": "scripts/pml_evaluation/extended/preserved_time_metric.py", "sha256": "648b9d6ee3915f39f61089b502053e5d6ff5f876f8dce75e9af329f72afb22d8" }, { "artifact_name": "scripts/pml_evaluation/extended/rank_extended.py", "sha256": "5ad141511e0256cebc644d82684a510e32b4587ce7ec24d75713fe724841efe4" }, { "artifact_name": "scripts/pml_evaluation/extended/rank_readout.py", "sha256": "5f097264e056d95b899be0c5a6fa84e2cdcf6861a5433a7aa3dbef0c3ec03a21" }, { "artifact_name": "scripts/pml_evaluation/extended/record_mos_reuse.py", "sha256": "4afb5a81cb03c436c94d1968c2b4454d7d7fe04468d99ea2009b27f61e01819a" }, { "artifact_name": "scripts/pml_evaluation/extended/run_edit_quality.py", "sha256": "2674be9ff053af075f70c31eba018987b7e570c62e7db399f004f673e0efbb30" }, { "artifact_name": "scripts/pml_evaluation/extended/run_edits.py", "sha256": "02327a7a8a4ee90f14484f2591f5fc1eb1ecf050bcc179487a2b44c247c4a830" }, { "artifact_name": "scripts/pml_evaluation/extended/run_preservation.py", "sha256": "2ff0eb162aca8ec812f397743be76584de12cabff43a75897842ed251de101cc" }, { "artifact_name": "scripts/pml_evaluation/extended/run_quantization.py", "sha256": "5d97b7a01aafa35a4ec6f87d69688413b324b7a186515657618b40c57c9c9122" }, { "artifact_name": "scripts/pml_evaluation/extended/run_quantized_quality.py", "sha256": "26b199103a5a2637932df9e3844860e36858b08c283b6399a1d650028c08020b" }, { "artifact_name": "scripts/pml_evaluation/extended/run_readout.py", "sha256": "5285f96bbdba6f667be0f9563a3873383ace1fdd69d4e2a879a3fb7f6d56b27e" }, { "artifact_name": "scripts/pml_evaluation/extended/run_train_features.py", "sha256": "a31607bbff4067bba106b65bc578df536ad4a066bbecc329442af3e3f9c31593" }, { "artifact_name": "scripts/pml_evaluation/extended/summarize_edits.py", "sha256": "7b6dc3de8d393eabc249b68f7db74e38f47ca65bcc7864e28a2d17871e00fec5" }, { "artifact_name": "scripts/pml_evaluation/extended/summarize_pitch_support.py", "sha256": "f087146e209c4eb6468b6c982b39a5720645f856dc80aee6f915f3bc0ecba951" }, { "artifact_name": "scripts/pml_evaluation/extended/summarize_preservation.py", "sha256": "d1d7a5bf299baf6e1342edc48f9bac2845593d31021219b4d5b36a1e6358ec77" }, { "artifact_name": "scripts/pml_evaluation/extended/summarize_quality.py", "sha256": "70ef09c1a8d56edbc2052a0210538690e046b8098b44dd3fc9326ab5db1013d4" }, { "artifact_name": "scripts/pml_evaluation/extended/summarize_quantization.py", "sha256": "9413742258238007a9988016c4bdc935a8650fdb9a9f0f8adc8b70b34fc4be6e" }, { "artifact_name": "scripts/pml_evaluation/extended/write_coverage.py", "sha256": "ab0733db95e154c8ae8737974bc428cf20036f554290226242f3d9446679302c" }, { "artifact_name": "scripts/pml_evaluation/portable_context.py", "sha256": "2eafd4d656111949891592b5905455eaf0b9956039ffdc85dafb107a8ba3d48b" }, { "artifact_name": "scripts/pml_evaluation/prepare_assets.py", "sha256": "b616cfabc48693ac260329e7994d74a057a88162ee1b0a0d8604f08500ec6f24" }, { "artifact_name": "scripts/pml_evaluation/prepare_inputs.py", "sha256": "98032ac7c518edad311f95100684c35988e6bf2f0ef8dbbba5f8cdb29da8295a" }, { "artifact_name": "scripts/pml_evaluation/run_asr.py", "sha256": "b7b16fe379335ecad13bd9ce64a8c02612ec92a03d13a441d7b3c2d370b4c1f8" }, { "artifact_name": "scripts/pml_evaluation/run_extended.py", "sha256": "ca3629c244a576ebb7dca732b44df81a871c991cdbd1ce9bb4e49cba48f317ce" }, { "artifact_name": "scripts/pml_evaluation/run_native.py", "sha256": "79ded7fc92c76a0c8d1cd14a22fe93f38071423ddbab37fa85b450ddcce233ea" }, { "artifact_name": "scripts/pml_evaluation/run_quality.py", "sha256": "8ea982cbe9d07fcd0f730b541567985f5512cadd9c6365237a6ae9edf989d11b" }, { "artifact_name": "scripts/pml_evaluation/run_seed_variation.py", "sha256": "cc820d7eba5ddf9377d92b5e4804f460af28d937efb0f528421398a0e840279d" }, { "artifact_name": "scripts/site_metric_comparison.py", "sha256": "9780c8ccdeb972b29a5d09a555426b5e961c13d9f7b9effe07e8d37e382e7687" }, { "artifact_name": "scripts/site_pml_report.py", "sha256": "09739f806a98505da657fdfafcd0513caaf2789de20ac180c509d6f5000b23cf" }, { "artifact_name": "scripts/site_preserved_reports.py", "sha256": "2cf501793ea73d73e4f3f26cf065d80955d17b5d809db7e9a0ea41df53eadb1a" }, { "artifact_name": "uv.lock", "sha256": "270f89c761dc253005840f140c520795684276c44635cce16707e8611b76a0d1" }, { "artifact_name": "vocoderbench/__init__.py", "sha256": "73738d70706824f85fdfac302306af505437a50b0ca4cdc68fd57aec211a308d" }, { "artifact_name": "vocoderbench/assessment/__init__.py", "sha256": "37126abeb4fe6f6ce927d264c42a1b2ab934eb9b11a25faa8165d1685c37f63c" }, { "artifact_name": "vocoderbench/assessment/__main__.py", "sha256": "ce535ce8695e480492699a7a859ad36daea0e8cf0f192f8dbbd76ed00891a2e0" }, { "artifact_name": "vocoderbench/assessment/acoustics.py", "sha256": "50737f8804f433b09737442e30c224a00e5f30f12583d38c4b7465b40d9669a3" }, { "artifact_name": "vocoderbench/assessment/common.py", "sha256": "ab938071273d5d475df871e521cba4ebb21e6d352c9e52d994c04cf71a957829" }, { "artifact_name": "vocoderbench/assessment/controls.py", "sha256": "8d632e7e95f5d674f034e865a4f532247cec41c548e27cc54dd75d09504653f3" }, { "artifact_name": "vocoderbench/assessment/curves.py", "sha256": "4a794d7d7a94666f298b46ff3c6b6ac943776660ad2ea4f1174fc922140006f1" }, { "artifact_name": "vocoderbench/assessment/probes.py", "sha256": "b1bced34227ce72b5e5fe640cf39c761172aeaa0ef9341829f19bdaf4a984f39" }, { "artifact_name": "vocoderbench/assessment/representation.py", "sha256": "84ea8e9de44e797c99066e36ce862a38759a96a4c8e0671f57c5a70c8c0b12b6" }, { "artifact_name": "vocoderbench/assessment/runner.py", "sha256": "44a522b612cbaae6f0dad91153cd315bc441964feb3b6faab24938a95fb3e677" }, { "artifact_name": "vocoderbench/audio_report.py", "sha256": "f46ff0cf69e7cb956ea94ee0c7013d44052c0400c5ffc64d7f9f90b4489aa980" }, { "artifact_name": "vocoderbench/audio_samples.py", "sha256": "b9b975c1f305cdaff47909b99bd2a3dce64996e90774d2a8305adc1c67ed6c17" }, { "artifact_name": "vocoderbench/benchmark.py", "sha256": "f53c6f2c40fcab2d6f8c5f9e0b1b637c9d21d7a3665b5571c733e111f62f1edc" }, { "artifact_name": "vocoderbench/benchmark_io.py", "sha256": "7bc0a304f61a3615e7d5c9cec91914a28803085f20ff52fed9d6d467dfcf8555" }, { "artifact_name": "vocoderbench/benchmark_pipeline.py", "sha256": "a049034dcd272bbafda47cbf5719cc6b1cd8dae6f80682200cd3fc7b90e522cd" }, { "artifact_name": "vocoderbench/benchmark_report.py", "sha256": "2a8061e8acc89dc4897d77efa206922720d6aa9843c340daea8ec78c2cb0b759" }, { "artifact_name": "vocoderbench/benchmark_summary.py", "sha256": "8221086fe08bfc70cbfb9308a3a01d6fea976afb7de86af8cbacd7999e8281da" }, { "artifact_name": "vocoderbench/build_world.py", "sha256": "7d29d81c9a48d633edfeae0425c5cc055bc0a10d230232a51eab5720b2cc6974" }, { "artifact_name": "vocoderbench/calibrate_pitch.py", "sha256": "282c917bf87c6f549f32467647802852907e4b0b88128eb8017157b84566adec" }, { "artifact_name": "vocoderbench/control.py", "sha256": "bc41e73ff34b5f9cb52280a6f2317c989abc53f8b15f8f82db5ee3102b4b58bd" }, { "artifact_name": "vocoderbench/corpus.py", "sha256": "bc14cee2c3dd1a651cf2e5f0e9fe871fbd3a712d1a90d7c18a18547a625802e8" }, { "artifact_name": "vocoderbench/extended_metrics.py", "sha256": "ab3537d2c8e5d568a0cd2fa8aa36309de09c3bce7cf2f849214331a22fe3624f" }, { "artifact_name": "vocoderbench/fixtures.py", "sha256": "2b2ba922ed611abc91a29f1d270c9a91c0743bc9a8a6a131c7dcd184fcf60dc9" }, { "artifact_name": "vocoderbench/logging_utils.py", "sha256": "c80eb5d9647516a4bc03980cb178f8b4bb130ac15877d1b3289797e7ad81e381" }, { "artifact_name": "vocoderbench/methods.py", "sha256": "f50d51c8559f41b8112338b681e6ec1f6382bfdda93b63f58aed9f7d8493bb77" }, { "artifact_name": "vocoderbench/metrics.py", "sha256": "b8767feb49494058bd9ec08fd95ae2fb50ce7897fcdd6cebe319ff8a0ac51f22" }, { "artifact_name": "vocoderbench/mos.py", "sha256": "fb632aa5440245525289962372db4f3eb007cd2b3b493b2d05ecccb878e85f03" }, { "artifact_name": "vocoderbench/mos_batch.py", "sha256": "c7a732dfe52b08fa155201a7134d2f5561a10b16d1866e9f0526f61b2b244f64" }, { "artifact_name": "vocoderbench/pml/__init__.py", "sha256": "5104033081c7bd776037b272740a608cb4dc255dc33eb632f0fbb5417d4223e3" }, { "artifact_name": "vocoderbench/pml/codec.py", "sha256": "4330edfe96d19d1a40647620fa6da5d9d2945fc8dc7b910751f8aaaae8c2c967" }, { "artifact_name": "vocoderbench/pml/fwcep.py", "sha256": "635564296ef5d1d1fcb889b2e4dc7ef5f90355568b7a926af7a0b80f089f4dad" }, { "artifact_name": "vocoderbench/pml/parameters.py", "sha256": "dcc3e45c28602354ee6b06a8622ed242c876fdfe90982630ef03deb997b0b2f5" }, { "artifact_name": "vocoderbench/pml/pml_analysis.py", "sha256": "9460b53c28d1fde3a99f66820de17b0d3f4bc18bec35ab43a349ca753b56708d" }, { "artifact_name": "vocoderbench/pml/pml_helpers.py", "sha256": "1d30cb14d521b8e5a681d4223e236223cf4a2378617f1561de06be503c93612f" }, { "artifact_name": "vocoderbench/pml/pml_pdd.py", "sha256": "d87f2f2cd3bdc16367e956eaa742502a6c2ebf509a4e63eb5e666c8e0ef7521c" }, { "artifact_name": "vocoderbench/pml/pml_synthesis.py", "sha256": "d0d35e3e3c7f25768fcb6052f4a7bd74734afe6514df8fc098b7a0a94f4157fc" }, { "artifact_name": "vocoderbench/pml/storage.py", "sha256": "6f9da667daca00da9e31f7be436b9bc412d2510a426359a43eeb0c6add154095" }, { "artifact_name": "vocoderbench/pml_cli.py", "sha256": "6ef4e7c783ec6ceb04ab9a9fb98e1688c7d0673b38a332b8f589ecb13438dd19" }, { "artifact_name": "vocoderbench/quality.py", "sha256": "f6a276d9ac2acdd345e49070c51d76bfa179e82f8459d8e15dbd1a050323d63d" }, { "artifact_name": "vocoderbench/quality_backends.py", "sha256": "1f36068515bc3e218f9eb149d7fc3b2b8e62b5eee804448ff30f01624dd23a48" }, { "artifact_name": "vocoderbench/quality_core.py", "sha256": "66bf6c08633424f544b2ad7eef2a74283c05d809aa9a59df8fc1ba0e5ab51f4b" }, { "artifact_name": "vocoderbench/quality_evidence.py", "sha256": "5e555c1747c41978bfb8d1a13948510e55898ab38714551d857196d2ac87052a" }, { "artifact_name": "vocoderbench/quality_inputs.py", "sha256": "5f71690e4daf12cf900029224c78232de34327c1ff4a23ee4376064367c40909" }, { "artifact_name": "vocoderbench/quality_reference_inputs.py", "sha256": "215c0967258c20af5b3267a00cb90aa1d0c7f22bb11e1efef98ad55711511a42" }, { "artifact_name": "vocoderbench/quality_report.py", "sha256": "aae925d97a6c990ad3318899a6b14995ecbf1cf6554041344bbed708fa51dad4" }, { "artifact_name": "vocoderbench/quality_similarity.py", "sha256": "32b537b5eb006834a55314a5ddb2f9faa7f5c9fa8d6a4faa626d892360d965f7" }, { "artifact_name": "vocoderbench/quality_visqol.py", "sha256": "908ea7cbf6b8202a46863c4c4b66d5a0e3adce817facd6898bd666a2e806099a" }, { "artifact_name": "vocoderbench/reference.py", "sha256": "3381b5141ad4eddaa25a90bcea778109f4040afaa3210f75f966522d716b46ad" }, { "artifact_name": "vocoderbench/references/__init__.py", "sha256": "d96022954c3cd493c7e6d7baf35e8bd33eafea192a9f9154c8470b7937d05732" }, { "artifact_name": "vocoderbench/references/adapters_dsp.py", "sha256": "ab5a0cf63085524842a0614d2ebc63adbe11521852f53bf9d561a131c37ab285" }, { "artifact_name": "vocoderbench/references/adapters_neural.py", "sha256": "cfe845d99cf0257556da99e5a1f950c9d9ac5ac490fb7d8e8cfb9b46f2c03a08" }, { "artifact_name": "vocoderbench/references/adapters_source.py", "sha256": "28b70e6680a3c8090fa5d16f696eae5c28d6b55bee5eee9e90ef4b80735f2b99" }, { "artifact_name": "vocoderbench/references/catalog.py", "sha256": "e6f98518adec991acce2aec3452ce333dbee93a3994f9d34e1fa4a2f8cbd8c63" }, { "artifact_name": "vocoderbench/references/common.py", "sha256": "faac86809af17b7a5a340a4ea8b9ba5b44c990375487b0cdfea94f59a78cabb5" }, { "artifact_name": "vocoderbench/references/current_base.py", "sha256": "edf5f29e6440897ce808911b195409a91568c733b836304832452111b3009b7d" }, { "artifact_name": "vocoderbench/references/evaluate.py", "sha256": "f37d09a45d8f801fb68d2e9eef31c0c2aca421a64bfdc42be4d233ccc815f917" }, { "artifact_name": "vocoderbench/references/evaluation_cache.py", "sha256": "4ff288c1f4f839db9004960333e1692c47f3e28cc19b2c8781a73e65cf453e0a" }, { "artifact_name": "vocoderbench/references/inputs.py", "sha256": "11acdfe0b4611075f3d232161008f098f59daf2a9345d7e1e75a4e1680ada1c8" }, { "artifact_name": "vocoderbench/references/integration.py", "sha256": "65e0917f8805b3f0ca85e463165354f2d2d1d530556f011ed21088fc61d6d2ac" }, { "artifact_name": "vocoderbench/references/native.py", "sha256": "a6581f389587b85f547ba66d068ce974d8fe6d08252500f0f28788a7014780fb" }, { "artifact_name": "vocoderbench/references/prepare.py", "sha256": "82092abb7f8e950b0c09b08568580627e6696796d054a716824e038ed80699d4" }, { "artifact_name": "vocoderbench/references/publish.py", "sha256": "618df5263723b4a94a653d3c636255af347071396a3f3c954e27a4290d37ae58" }, { "artifact_name": "vocoderbench/references/report.py", "sha256": "011afb9c6453d9b215f7bc687d9e2dbafbdaddfdd9864e8925ce01ab9cb88fcc" }, { "artifact_name": "vocoderbench/references/report_formal.py", "sha256": "b7054982d383dfc7e6aa88f27fc56ef30d0edd057b2d48d826037dc15f0e6de0" }, { "artifact_name": "vocoderbench/references/report_metrics.py", "sha256": "f4e3e1424b091433d34aae77383cbb51c6063e7803f3c5381375a6fd8b5f33f3" }, { "artifact_name": "vocoderbench/references/report_summary.py", "sha256": "c92f65aca61793b5907470dacd31ef0afb2d6ca310878417b3e4ce0beb5c2bd9" }, { "artifact_name": "vocoderbench/references/worker.py", "sha256": "9f2ad43a514b6fcaa048365908f44a73b55d0a7d37974cb6a88777cd6b80a0d7" }, { "artifact_name": "vocoderbench/report.py", "sha256": "659d8406e8c1d09e29e120d6f0ef13ddc3047565915c125588a2704fb25c6737" }, { "artifact_name": "vocoderbench/run.py", "sha256": "c9e846e0e9e39fe25fec9daa4cb5709b31734efd2e44e1614ad0722c99fd2cd2" }, { "artifact_name": "vocoderbench/site_summary.py", "sha256": "e18f4ab57f8e809006a4758cd62b054736b6208768474ff319b10a12b4cffef2" }, { "artifact_name": "vocoderbench/supplement_report.py", "sha256": "9146b635745c6bb10c733c090696c564d463180507809588ac377ef682c668ad" }, { "artifact_name": "vocoderbench/temporal_metrics.py", "sha256": "ce21b6a4767071a7b98ea0cdbfe54bdba15b596bf8d6b6127281a97095e7be91" }, { "artifact_name": "vocoderbench/world.py", "sha256": "d14558eed772e07031a9183f57facc55acb1c27deccb3dd4fa098ffcbe0cde3a" } ], "scope": "Frozen evaluator source and asset identities only; implementation code and model weights are not included." }, "evidence_sha256": { "metric_summary": "1bee17075fc78e939cb4e36fe10ff23e5be87ff12585bf9758f8e6c0136b7f14", "metric_completion_receipt": "ff07641c19cb6a6ef2ab4e8807d4b6abc50a2a49ff82258a74c1b9954b6df142", "model_initialization_record": "a0bc15324617145cc8e3361a41d2b6797be6d87663e7b0dba086b2a2bce5fc3e", "B_waveform_generation_receipt": "f0fb857bda6041d7a174c67266a643015347a981c14938dcfd29cbb2ad14337b", "A_B_waveform_identity_record": "6a2b3fed8406c8ece9f87d04c9e4365f595f6ffc7d5a3d1f512b2d185cbd0fab", "initial_preliminary_observations_json": "4dac5af343261e386aa12c8d5b32cbc968c6c8e1b6d0f179dcf205da74e648b1", "initial_reproducibility_json": "994f3b6a7a4de3949de559ea2539bb3c87354c0da79d24f38df5b8b7ed07d0e9" }, "limits": [ "Predicted MOS is not a human-rating result; calibration for these vocoders has not been performed.", "YIN measures relative edit tracking against the same variant and seed native synthesis. Native absolute F0 error is unmeasured; cross-variant support sets and voiced-frame denominators may differ.", "Recognition is available for only two of the five recordings. The remaining three have no prepared official excerpt-specific transcript; no ASR output was substituted as ground truth.", "Matching CER/WER does not establish waveform, transcript, phoneme-onset or perceptual-quality identity.", "An absent threshold violation is not proof of no harm. The original acoustic warnings are preserved.", "All observations share five recordings and four speakers. Repeated seeds and executions are not independent samples; the means are descriptive.", "Source-periodic AM was not separated. No adoption or formal 120-recording evaluation follows from this addendum." ], "decision": { "baseline": "PML", "RAW_FULL_adopted": false, "RAW_SHAPE_adopted": false, "source_periodic_AM_separation": "not_attempted", "formal_120_evaluation": "not_run", "expansion": "on_hold", "necessary_observations_missing": true }, "public_projection": { "included": "All 90 quality scalar records with their missing-value reasons and support counts, all 60 saved paired comparisons, fixed model and source hashes, and descriptive aggregates.", "omitted": [ "framewise F0/time arrays", "ASR transcript text", "implementation code", "audio", "model weights" ], "array_omission": "Each full per-condition metric JSON is identified by SHA-256. The three F0 frame arrays are omitted from the public scalar projection; all their saved scalar summaries and support counts remain.", "attribution": "The LibriSpeech attribution and source modifications are documented in README-ja.md and reproducibility.json." } }