Spaces:
Running
Running
| { | |
| "schema_version": 1, | |
| "catalog_fingerprint": "746b255476160936117a35a21dc2ded1b803bb26bf6046e16ed87c98281eb8ff", | |
| "selection_sha256": "ac2de6a310092df70ba262743459ca7e2f837db1ab263d793f5a3746e88da5d5", | |
| "counts": { | |
| "text": 6, | |
| "emotion": 5, | |
| "prosody": 5, | |
| "pause": 5, | |
| "multi": 5, | |
| "noise": 2, | |
| "multi_speaker": 1 | |
| }, | |
| "sample_count": 29, | |
| "audio_count": 136, | |
| "audio_bytes": 74207080, | |
| "figures": [ | |
| { | |
| "path": "assets/dots.tts.edit.drawio.png", | |
| "bytes": 170384, | |
| "sha256": "c9bd94dd9824d71a7ed6106fee95c006cb19ad29a56269a4c1550ea0f9c31aaa", | |
| "drive_folder_id": "1taUnf-RzYIS4Z-o8a8szzwHA_vgURaCg", | |
| "drive_file_id": "1eBP9tGQV7TP8kmm_D-pakCcpqbZk0xq0", | |
| "drive_file_name": "dots.tts.edit.drawio.png", | |
| "drive_modified_time": "2026-07-30T08:24:02.201Z" | |
| }, | |
| { | |
| "path": "assets/doteb_distribution.svg", | |
| "bytes": 20669, | |
| "sha256": "939c308d6533a13e2ba66c02f20beb4af3ceb5649d5474248ec8b25e3b40606c" | |
| }, | |
| { | |
| "path": "assets/radar_models.png", | |
| "bytes": 1076040, | |
| "sha256": "980f2e8b574995e82c786e2c39b462c598a09e1a8f6a29a31b1ac6fa0a851bc9" | |
| } | |
| ], | |
| "audio": [ | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0032/source.wav", | |
| "bytes": 257358, | |
| "sha256": "7432934a8228c4a89123ae2d4a0b64c96da38223921d2daafc20eb6d992c3295", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 128640, | |
| "duration_seconds": 5.36, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0032/ours_45k.wav", | |
| "bytes": 568364, | |
| "sha256": "c2297d873f1260591bf65b36acb4ee4f85ea76833d8cbeaedbd978e69455da3d", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 284160, | |
| "duration_seconds": 5.92, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0032/qwen3_omni.wav", | |
| "bytes": 248534, | |
| "sha256": "a2f652e09c8a68e30cd749dce1b49511702a02b63b8720c966cc279010839191", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 124245, | |
| "duration_seconds": 5.176875, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0032/mimo_audio_instruct.wav", | |
| "bytes": 314924, | |
| "sha256": "9557a6d85468c038674896ca4b8edd8196cad52bba817f6bb2e17d0810ca9517", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 157440, | |
| "duration_seconds": 6.56, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0032/expert_data_pipeline.wav", | |
| "bytes": 293470, | |
| "sha256": "2c0bfddbce09c7b4a2289dc3627852156a30a149e248573e7fe36a1f80844186", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 146713, | |
| "duration_seconds": 6.113042, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0057/source.wav", | |
| "bytes": 923606, | |
| "sha256": "4f59982b599a13b0655400199d27fc832418ed69e27816de4ddb2ed7b64f0cac", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 461764, | |
| "duration_seconds": 19.240167, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0057/ours_45k.wav", | |
| "bytes": 1797164, | |
| "sha256": "924732b57e6442bd741bb462b2096bf516d7cbee5d92c713eb1c6313df0cbb10", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 898560, | |
| "duration_seconds": 18.72, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0057/qwen3_omni.wav", | |
| "bytes": 943574, | |
| "sha256": "5a5e1f2a5cee99560395f61046d35e178d598c925a90a8495f5e999c1a02424f", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 471765, | |
| "duration_seconds": 19.656875, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0057/mimo_audio_instruct.wav", | |
| "bytes": 545324, | |
| "sha256": "c9ce69a4397388c938e7641daeb22fa0c25e1f6e9659043bdc227fe3a237ae76", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 272640, | |
| "duration_seconds": 11.36, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_en_0057/expert_data_pipeline.wav", | |
| "bytes": 884500, | |
| "sha256": "374dd9bf71cb69a55824078383451d485803d276735d48b3785054955feb7236", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 442228, | |
| "duration_seconds": 18.426167, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0052/source.wav", | |
| "bytes": 509262, | |
| "sha256": "3c4309666fd189760db1d8e8ab3933004dd0c0cc69715a1a28f52ba3fcef111f", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 254592, | |
| "duration_seconds": 5.304, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0052/ours_45k.wav", | |
| "bytes": 629804, | |
| "sha256": "66a66d95c169da56da30efaade292bf92dc7f25b07b2dbf92904e01e1e487c94", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 314880, | |
| "duration_seconds": 6.56, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0052/qwen3_omni.wav", | |
| "bytes": 133334, | |
| "sha256": "410bbe23ec6bf3983afde90e8c094f82525a3e1922460bf6adffd3bde861da7c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 66645, | |
| "duration_seconds": 2.776875, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0052/mimo_audio_instruct.wav", | |
| "bytes": 422444, | |
| "sha256": "a22cdc31ae5cb4c9870c6421937f870f71837a2ae9752c73384fa20ff8b8ab7b", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 211200, | |
| "duration_seconds": 8.8, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0052/expert_data_pipeline.wav", | |
| "bytes": 742370, | |
| "sha256": "3fc4dc49576934bf6b146b77fa88f2591cbb5e893b749929abaacadd624e9215", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 371163, | |
| "duration_seconds": 7.732563, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0075/source.wav", | |
| "bytes": 357198, | |
| "sha256": "f22c0213a519bed34b8a47f7662e2c155d17dd0b5c2df4eecff02e851d529fb4", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 178560, | |
| "duration_seconds": 11.16, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0075/ours_45k.wav", | |
| "bytes": 967724, | |
| "sha256": "b4aad9ed9834f46cddc2680f50fd9b492d0c0110d7093c741a559add4b323354", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 483840, | |
| "duration_seconds": 10.08, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0075/qwen3_omni.wav", | |
| "bytes": 655574, | |
| "sha256": "c2943025bd83cb9a14e8b4fba5bbcc87a4d7628cb9a77bbe19273e63e7183d3b", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 327765, | |
| "duration_seconds": 13.656875, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0075/mimo_audio_instruct.wav", | |
| "bytes": 491564, | |
| "sha256": "9ee57b46cfeebc9a3d0e7bbecf9bdb2d45f21b13fbb133064a2b868aeb3ee4ea", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 245760, | |
| "duration_seconds": 10.24, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__easy-0727__0727_zh_0075/expert_data_pipeline.wav", | |
| "bytes": 356340, | |
| "sha256": "c486b67594019ce87c50b30529200cae47aeac4cf57d0f747ee2cfedd518d797", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 178148, | |
| "duration_seconds": 11.13425, | |
| "benchmark": "text", | |
| "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_en_0107/source.wav", | |
| "bytes": 497358, | |
| "sha256": "5bca656e973aaa6c342b20dbd04316883958b257e0435a5ac3fdcbc22753313c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 248640, | |
| "duration_seconds": 10.36, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_en_0107/ours_45k.wav", | |
| "bytes": 169004, | |
| "sha256": "b660b2c64d7d9ed051f62d7d1a7d8a3750581f6856f4163be7d4e5d75e518e4a", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 84480, | |
| "duration_seconds": 1.76, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_en_0107/qwen3_omni.wav", | |
| "bytes": 56534, | |
| "sha256": "92496f1a91d1583215aff0fd0440c9b308c1a731b8587dd697f4811d3dd68af5", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 28245, | |
| "duration_seconds": 1.176875, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_en_0107/mimo_audio_instruct.wav", | |
| "bytes": 92204, | |
| "sha256": "6ea56f4fc6b83ae7ed04d15befb6976a7db6c43681ac58982b349a5a36d1f9d2", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 46080, | |
| "duration_seconds": 1.92, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_en_0107/expert_data_pipeline.wav", | |
| "bytes": 128842, | |
| "sha256": "904a5efef6b3d0ca29b40d051924c46a95681096c17ec7fa92dc822ab9f3791f", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 64399, | |
| "duration_seconds": 2.683292, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/source.wav", | |
| "bytes": 1011534, | |
| "sha256": "64686ceb28f92be6245769397e256a251a9ce6cce295fad6530a8e8c93af389b", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 505728, | |
| "duration_seconds": 10.536, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/ours_45k.wav", | |
| "bytes": 199724, | |
| "sha256": "dd3978f059c5baa0bc394cb4eb4dd508cece06bb41d568ab19df942068920682", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 99840, | |
| "duration_seconds": 2.08, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/qwen3_omni.wav", | |
| "bytes": 41174, | |
| "sha256": "bb75f2de95a16b03d9ac3ecb8232166951edd8f102df8494985c30ba81eb1a68", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 20565, | |
| "duration_seconds": 0.856875, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/mimo_audio_instruct.wav", | |
| "bytes": 69164, | |
| "sha256": "b8f78a0d874b005b9d1fb7e255dd24a082a7ee349b7396b9205599cb64ead869", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 34560, | |
| "duration_seconds": 1.44, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/expert_data_pipeline.wav", | |
| "bytes": 360562, | |
| "sha256": "fae3644fc6b36720c6b319155f83e5abc94c0e9cc0341fc0aa8616993cc1c1cf", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 180259, | |
| "duration_seconds": 3.755396, | |
| "benchmark": "text", | |
| "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/source.wav", | |
| "bytes": 591412, | |
| "sha256": "829a755290ead5300e9d88132191ffa7f6533a3c664866bf5cb423d875bb356e", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 295684, | |
| "duration_seconds": 12.320167, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/ours_45k.wav", | |
| "bytes": 1182764, | |
| "sha256": "2d87ecf667862d47d2de7effb819028483d2feeac22e8eadbce70ea13cd6f314", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 591360, | |
| "duration_seconds": 12.32, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/step_audio_editx.wav", | |
| "bytes": 414764, | |
| "sha256": "e8b4562ff4fee4a03a9aa4ecc03b89410d89850a038de783c2324236d016a74c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 207360, | |
| "duration_seconds": 8.64, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/ming_uniaudio.wav", | |
| "bytes": 387278, | |
| "sha256": "c0d47028fdf9aaeec7e6041b47ae5cd1d53142831b9447a0699f53bdca9ef102", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 193600, | |
| "duration_seconds": 12.1, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/expert_data_pipeline.wav", | |
| "bytes": 579892, | |
| "sha256": "b3e4b55c18d0e2aa732c14d4f005b39096d8b6ad9aec226b8c22f8c4af8c370d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 289924, | |
| "duration_seconds": 12.080167, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/source.wav", | |
| "bytes": 435884, | |
| "sha256": "170faade763c15fd12b0dfd7d8843460877e39b2b05ef40821740171ea10b9d0", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 217920, | |
| "duration_seconds": 9.08, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/ours_45k.wav", | |
| "bytes": 921644, | |
| "sha256": "9798bd43e4adbfdc9fd58bf940f9246752cd81f56c3edfb3ae3ea41dc058a4ea", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 460800, | |
| "duration_seconds": 9.6, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/step_audio_editx.wav", | |
| "bytes": 385964, | |
| "sha256": "d252df8bcfe88d5a4ec78a68eed2be44556891218768ee36d2b62ae77f3eab2a", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 192960, | |
| "duration_seconds": 8.04, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/ming_uniaudio.wav", | |
| "bytes": 326478, | |
| "sha256": "54addb0021a1e07feeac8dff30e297d152a29d447740d5a81c9c3db369bda13b", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 163200, | |
| "duration_seconds": 10.2, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/expert_data_pipeline.wav", | |
| "bytes": 505004, | |
| "sha256": "4d7c8b5070aad0133011be36bdc53efd5f80620ca391d0498b43d4dfbb2d8ad9", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 252480, | |
| "duration_seconds": 10.52, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/source.wav", | |
| "bytes": 912428, | |
| "sha256": "15463a828fc509f73e8dde565f6b9d9ea7d0953fc5c0093fe88501cdb4f9b0b0", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 456192, | |
| "duration_seconds": 9.504, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/ours_45k.wav", | |
| "bytes": 967724, | |
| "sha256": "a764f101ee537acb8b0373f6e857bb6308c56826acccb94271fa8e70b1f75a86", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 483840, | |
| "duration_seconds": 10.08, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/step_audio_editx.wav", | |
| "bytes": 529964, | |
| "sha256": "077bf1ee11783ec24cc830f89fa2484e01f398b2764b9b6d5adbe4cf37488d4f", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 264960, | |
| "duration_seconds": 11.04, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/ming_uniaudio.wav", | |
| "bytes": 156878, | |
| "sha256": "716f33aa37ab047336a0c48fcf8c6a3a76f453a0982032e7727ac2adce235bc1", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 78400, | |
| "duration_seconds": 4.9, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/expert_data_pipeline.wav", | |
| "bytes": 579116, | |
| "sha256": "a195a27ecafda2e8daf5e5af19f92b9c0a082eb16225017b46c4e441adf1e00f", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 289536, | |
| "duration_seconds": 12.064, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/source.wav", | |
| "bytes": 753452, | |
| "sha256": "85aada0c5e0de74544c5dd5b28b0351168b6ad2ded59004510f6cf800943de3f", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 376704, | |
| "duration_seconds": 7.848, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/ours_45k.wav", | |
| "bytes": 691244, | |
| "sha256": "832fe2b0a76ec821d648db9dcff5ea8d3da05851353152260e869425f30f6e7d", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 345600, | |
| "duration_seconds": 7.2, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/step_audio_editx.wav", | |
| "bytes": 668204, | |
| "sha256": "5fc9e6df5d105cc7452d2db326a1c81c37762841ec1fdecf8e5d2f1b678d9cd5", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 334080, | |
| "duration_seconds": 13.92, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/ming_uniaudio.wav", | |
| "bytes": 220878, | |
| "sha256": "dfeca11851524a1d35140a8dacd307bf97aeb3a558336b552d26846584dd35db", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 110400, | |
| "duration_seconds": 6.9, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/expert_data_pipeline.wav", | |
| "bytes": 468908, | |
| "sha256": "931875a7692475434b8935c81581d6a3888725d410a9dacf7d8e361a5f232fbd", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 234432, | |
| "duration_seconds": 9.768, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/source.wav", | |
| "bytes": 368684, | |
| "sha256": "af31953320e60245be7102c81b563726b5fba2d68476c41462664f9868dd6f93", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 184320, | |
| "duration_seconds": 7.68, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/ours_45k.wav", | |
| "bytes": 844844, | |
| "sha256": "e2429658a0468140a76e4c4324e97fbcea91a5576e13c0ef675a9d682272287c", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 422400, | |
| "duration_seconds": 8.8, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/step_audio_editx.wav", | |
| "bytes": 195884, | |
| "sha256": "662a5428c4b6cfdb051867defd7be1a3af835eee4678a2bddc6e5629155849b3", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 97920, | |
| "duration_seconds": 4.08, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/ming_uniaudio.wav", | |
| "bytes": 230478, | |
| "sha256": "01999c5a1de5e346b983f6fd7c73527f66d5b69b2cdb877f3daa7c44e3f6aaac", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 115200, | |
| "duration_seconds": 7.2, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/expert_data_pipeline.wav", | |
| "bytes": 451440, | |
| "sha256": "3928c434dc2908c3c065b3c5614c0074d3a749afe343a5b2001f955b1a451cc8", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 225698, | |
| "duration_seconds": 9.404083, | |
| "benchmark": "emotion", | |
| "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/source.wav", | |
| "bytes": 245804, | |
| "sha256": "1db7cbee09f11b4d2d454c6391c8fc2b9fe47131c9951b3fe2d7b08ce5b3f519", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 122880, | |
| "duration_seconds": 5.12, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/ours_45k.wav", | |
| "bytes": 537644, | |
| "sha256": "6d828a31c7c0c853ee9da2bbc183e968d848a105ea2caaaa6c55632b28f4e980", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 268800, | |
| "duration_seconds": 5.6, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/ming_uniaudio.wav", | |
| "bytes": 124878, | |
| "sha256": "923a8c958d716d837c98bf045901eff595fdc80ceaf575a2350dad35f82c9965", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 62400, | |
| "duration_seconds": 3.9, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/qwen3_omni.wav", | |
| "bytes": 210134, | |
| "sha256": "9a353a1af8b0bc03871a4f271985c10cf6bad90d084830a1701f76dc254c5676", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 105045, | |
| "duration_seconds": 4.376875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/expert_data_pipeline.wav", | |
| "bytes": 249924, | |
| "sha256": "5da6597bd1538feec9e07828b92f0e41283e4305b9f1266a856d415e7a630723", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 124940, | |
| "duration_seconds": 5.205833, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/source.wav", | |
| "bytes": 188204, | |
| "sha256": "98e62e7b841e48aacff5826358900c95c659a915f3e9dc594f8b70417e059045", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 94080, | |
| "duration_seconds": 3.92, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/ours_45k.wav", | |
| "bytes": 414764, | |
| "sha256": "0a68b745eb2e2d556a47180ff826ca47855af3429e89eddf3ab68b08fb97f7e6", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 207360, | |
| "duration_seconds": 4.32, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/ming_uniaudio.wav", | |
| "bytes": 54478, | |
| "sha256": "b0cd856b19c127a30a077ee3ee0b71e9cd17f654ba9a1eb95b77e010c66898f7", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 27200, | |
| "duration_seconds": 1.7, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/qwen3_omni.wav", | |
| "bytes": 194774, | |
| "sha256": "a5953eaf77b4f0c8cdf76981e902afebab6315080da12444e965ba3c344faa6c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 97365, | |
| "duration_seconds": 4.056875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/expert_data_pipeline.wav", | |
| "bytes": 188204, | |
| "sha256": "ba945ab06a47d49d04d08d51a5af03d57d9678b7b07006e3fd5047787707f9ad", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 94080, | |
| "duration_seconds": 3.92, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/source.wav", | |
| "bytes": 290934, | |
| "sha256": "f1a1e0a79115d9385994347bfaed1bacce78840ddddd9af70992260950025e06", | |
| "channels": 1, | |
| "sample_rate": 44100, | |
| "sample_width_bytes": 2, | |
| "frames": 145408, | |
| "duration_seconds": 3.297234, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/ours_45k.wav", | |
| "bytes": 337964, | |
| "sha256": "d7483c52e46e7383f8c432706cf948fed13437053ba130c26069e0fe2932199a", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 168960, | |
| "duration_seconds": 3.52, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/ming_uniaudio.wav", | |
| "bytes": 54478, | |
| "sha256": "b0f006951d6d9068d2aa92f08a4bc473f0abac49bfbf0655d714d5e727bd5b1a", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 27200, | |
| "duration_seconds": 1.7, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/qwen3_omni.wav", | |
| "bytes": 114134, | |
| "sha256": "df6e70c94143b4fe41cfecf6e5a3bf770d9f3939978cc113f7b1fb18f3ba2592", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 57045, | |
| "duration_seconds": 2.376875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/expert_data_pipeline.wav", | |
| "bytes": 158312, | |
| "sha256": "049d46bdf3ee875ad4fe2690b79216294ba8c902c17e4bd720dcd1ef90cce024", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 79134, | |
| "duration_seconds": 3.29725, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/source.wav", | |
| "bytes": 898612, | |
| "sha256": "e5eb50bde7347710b4bc09d2d43c37d6ac490a8e7dd5a83cd95091e2723d75b5", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 449284, | |
| "duration_seconds": 18.720167, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/ours_45k.wav", | |
| "bytes": 2042924, | |
| "sha256": "5f3f23fb0717309d4d56ccd74e474e6c9d824306e6e85917a1a63f2b07ada734", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 1021440, | |
| "duration_seconds": 21.28, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/ming_uniaudio.wav", | |
| "bytes": 601678, | |
| "sha256": "b6ef472c85b18be4dceddd87ea5a5de087d07adc85c018519c1898937f0a4994", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 300800, | |
| "duration_seconds": 18.8, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/qwen3_omni.wav", | |
| "bytes": 659414, | |
| "sha256": "a5b92e7516e99e4b767c29aa27304a110c32bc6e9b2580b1f637334ef8a37f94", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 329685, | |
| "duration_seconds": 13.736875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/expert_data_pipeline.wav", | |
| "bytes": 1017584, | |
| "sha256": "ab53eddfc8a970a8ab7d934844a46f62468518f807993dba80adb419314c0062", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 508770, | |
| "duration_seconds": 21.19875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/source.wav", | |
| "bytes": 557296, | |
| "sha256": "620c9c909835c01014089f5347f696af59e8775a059dabc15f79019223b792d8", | |
| "channels": 1, | |
| "sample_rate": 44100, | |
| "sample_width_bytes": 2, | |
| "frames": 278626, | |
| "duration_seconds": 6.31805, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/ours_45k.wav", | |
| "bytes": 476204, | |
| "sha256": "46d2256464c7c0dcb2abba1994b231d2128a64fdf9e9398a218b946edb95364b", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 238080, | |
| "duration_seconds": 4.96, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/ming_uniaudio.wav", | |
| "bytes": 195278, | |
| "sha256": "e0b3091d995df388cc6beb23a1c6fec7af598e6e27aef7a04df3f8acc1b121d0", | |
| "channels": 1, | |
| "sample_rate": 16000, | |
| "sample_width_bytes": 2, | |
| "frames": 97600, | |
| "duration_seconds": 6.1, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", | |
| "role": "ming_uniaudio" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/qwen3_omni.wav", | |
| "bytes": 94934, | |
| "sha256": "91ea4e7257138093e865891cf71decf6a24435ab9adcd00a2e231d113ff07f38", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 47445, | |
| "duration_seconds": 1.976875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/expert_data_pipeline.wav", | |
| "bytes": 261638, | |
| "sha256": "5cf8ad0f4b02957aae972908e893fe3979bc30d1cb3c282391d9b087c8e48ed8", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 130797, | |
| "duration_seconds": 5.449875, | |
| "benchmark": "prosody", | |
| "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/source.wav", | |
| "bytes": 466638, | |
| "sha256": "3aeb98e4fb92c0ae315a71711b6e0a241128a23457f5b37342c16e6fa821b486", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 233280, | |
| "duration_seconds": 9.72, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/ours_45k.wav", | |
| "bytes": 1044524, | |
| "sha256": "e58911d19235412f26d6a0d135dcf25ee786ba1d01068dea7c30963afaed25a9", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 522240, | |
| "duration_seconds": 10.88, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/mimo_audio_instruct.wav", | |
| "bytes": 476204, | |
| "sha256": "e1ffe12d74bf48565190c26c98a68acce9341e50e6d96d38084679d339fb052c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 238080, | |
| "duration_seconds": 9.92, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/qwen3_omni.wav", | |
| "bytes": 390614, | |
| "sha256": "6bd8c691a528f013324f428fd721836dfaef7fec6dbd04a8bae97ab7845bdad9", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 195285, | |
| "duration_seconds": 8.136875, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/expert_data_pipeline.wav", | |
| "bytes": 505004, | |
| "sha256": "de6d867b0ea58c9258fabdb80fa7c03d48a9a7be91d91cd932af3093463df309", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 252480, | |
| "duration_seconds": 10.52, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/source.wav", | |
| "bytes": 370638, | |
| "sha256": "65a937918451344f824d41c709a810947866ecec8f0f39a1f60cfbd16aad6765", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 185280, | |
| "duration_seconds": 3.86, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/ours_45k.wav", | |
| "bytes": 476204, | |
| "sha256": "cec691e73b578f1a0d449eab15f1d235be9367e7a48dfb47bae8e4e4191df9be", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 238080, | |
| "duration_seconds": 4.96, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/mimo_audio_instruct.wav", | |
| "bytes": 230444, | |
| "sha256": "9a3c80c350e4c06a287222752c0750b700cbb4b1b450acc49fb00d54c129ec40", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 115200, | |
| "duration_seconds": 4.8, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/qwen3_omni.wav", | |
| "bytes": 164054, | |
| "sha256": "4035da26693481c39e306a292ca76b121fb30604db1626fac07ef8dccce0241d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 82005, | |
| "duration_seconds": 3.416875, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/expert_data_pipeline.wav", | |
| "bytes": 447404, | |
| "sha256": "0e008749962e83ba18f417e9234f776d34f1217e55ab183baa27f7ea0a85efff", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 223680, | |
| "duration_seconds": 4.66, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/source.wav", | |
| "bytes": 615712, | |
| "sha256": "c3ffdfdf2ab0fbd4941b07d677508246b2f6faeaa36abba311ccc2e467947648", | |
| "channels": 1, | |
| "sample_rate": 44100, | |
| "sample_width_bytes": 2, | |
| "frames": 307817, | |
| "duration_seconds": 6.979977, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/ours_45k.wav", | |
| "bytes": 645164, | |
| "sha256": "3be8c5ed489861d6a341e749fbe79d1df0e0d359d89135592cf87154bc95a261", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 322560, | |
| "duration_seconds": 6.72, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/mimo_audio_instruct.wav", | |
| "bytes": 299564, | |
| "sha256": "85d363b354d723c7df54322699a7f522cf9fca27e832e97c34b1ac196fdde4bc", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 149760, | |
| "duration_seconds": 6.24, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/qwen3_omni.wav", | |
| "bytes": 210134, | |
| "sha256": "c1ad835d5a4706b436e1a7bc65fec16eeeb58e7c84650406f6c5883ed20641b1", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 105045, | |
| "duration_seconds": 4.376875, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/expert_data_pipeline.wav", | |
| "bytes": 650958, | |
| "sha256": "9dfd3a090efa4a16fb5dcc6aa724ffff810dbc0f5c8987c2920ced32d69c0c04", | |
| "channels": 1, | |
| "sample_rate": 44100, | |
| "sample_width_bytes": 2, | |
| "frames": 325457, | |
| "duration_seconds": 7.379977, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/source.wav", | |
| "bytes": 243918, | |
| "sha256": "c92b03c64a7526617add680d745e3659b929374a6b8e9a7bf82592af3e40f592", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 121920, | |
| "duration_seconds": 5.08, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/ours_45k.wav", | |
| "bytes": 476204, | |
| "sha256": "6b8568e16bfdd9da229c102cd24d63aa90e2295f97c408802c2abfe8d3414dee", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 238080, | |
| "duration_seconds": 4.96, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/mimo_audio_instruct.wav", | |
| "bytes": 330284, | |
| "sha256": "18f570b3ba32f966b122acb3341a868f397c9210c0c2e1baee4f02bdfd73762b", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 165120, | |
| "duration_seconds": 6.88, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/qwen3_omni.wav", | |
| "bytes": 263894, | |
| "sha256": "b772e54856bd93b986f2f6245ce6df3f444d1ec42c0f7f0ce01967185b00ca53", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 131925, | |
| "duration_seconds": 5.496875, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/expert_data_pipeline.wav", | |
| "bytes": 219884, | |
| "sha256": "f0eee3a657add295a568aca91422703519de5c8a71b1887cdd3c860e6029570d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 109920, | |
| "duration_seconds": 4.58, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/source.wav", | |
| "bytes": 576086, | |
| "sha256": "185f35e2e29074f77580d50c5318d878802785ecfe2b80f5c25d4bb69c4c9d7e", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 288004, | |
| "duration_seconds": 12.000167, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/ours_45k.wav", | |
| "bytes": 1136684, | |
| "sha256": "4e97cfca7b47d8a6ce83e3e134ff21d1315a1e58ef3a46c2ec793813958dd4a9", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 568320, | |
| "duration_seconds": 11.84, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/mimo_audio_instruct.wav", | |
| "bytes": 384044, | |
| "sha256": "76003f96103acfb4db5424034ffeef88db7004b9e00337541a338a898a59f696", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 192000, | |
| "duration_seconds": 8.0, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", | |
| "role": "mimo_audio_instruct" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/qwen3_omni.wav", | |
| "bytes": 432854, | |
| "sha256": "4334f6a1062810181110f30ffffe57aa33f5ae441680b623ecdf0c933423f323", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 216405, | |
| "duration_seconds": 9.016875, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/expert_data_pipeline.wav", | |
| "bytes": 547252, | |
| "sha256": "72894de9b7df59e90520c20fb6a233b190cec890489e45e510d0b83f9e3e5861", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 273604, | |
| "duration_seconds": 11.400167, | |
| "benchmark": "pause", | |
| "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/source.wav", | |
| "bytes": 750806, | |
| "sha256": "59f3f0adaaf57f0eaafa31db38c41e97973584b766b4e865cebad372017af0ec", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 375364, | |
| "duration_seconds": 15.640167, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/ours_45k.wav", | |
| "bytes": 1489964, | |
| "sha256": "651f7c55d77e4343691385a5c8da45fe7cc9dddbcda4835c6120e9af152889ea", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 744960, | |
| "duration_seconds": 15.52, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/qwen3_omni.wav", | |
| "bytes": 640214, | |
| "sha256": "3db54f1153ea9ca92bca47b676fca3b0cf1caa21aeca63236e9bff57fe5b3eb6", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 320085, | |
| "duration_seconds": 13.336875, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/step_audio_editx.wav", | |
| "bytes": 933164, | |
| "sha256": "3cf624b118e5f6b6904dd8af68d059fd8460c99336dd594eb3b6c21eb679d589", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 466560, | |
| "duration_seconds": 19.44, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/expert_data_pipeline.wav", | |
| "bytes": 719852, | |
| "sha256": "4364fbf52e459c5d2c478ffc73cad6ce68f30b9fc66a93032e40bacffcc0ea27", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 359904, | |
| "duration_seconds": 14.996, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/source.wav", | |
| "bytes": 702806, | |
| "sha256": "b8b536341bd1d6c425995bc3b49ab67c9818cf5aa88a57544129378929c55a65", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 351364, | |
| "duration_seconds": 14.640167, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/ours_45k.wav", | |
| "bytes": 1413164, | |
| "sha256": "19dfa6691298b7151d080cc20819b78091143a8ae7fe864ba7b71e77cb64e77f", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 706560, | |
| "duration_seconds": 14.72, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/qwen3_omni.wav", | |
| "bytes": 536534, | |
| "sha256": "e184619fa3d21cdaca6ce07768634d90fbb2e7cf3461dc9ef8abadd63bc2cdcc", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 268245, | |
| "duration_seconds": 11.176875, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/step_audio_editx.wav", | |
| "bytes": 1526444, | |
| "sha256": "74524f35fb878fc316e409419f826225f18019f25b68017c96c6efae3bc3f34e", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 763200, | |
| "duration_seconds": 31.8, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/expert_data_pipeline.wav", | |
| "bytes": 747892, | |
| "sha256": "a854ac9f58fd4e994719a280c0bf54a6ce6116987adbcf59428f1ade297e3f37", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 373924, | |
| "duration_seconds": 15.580167, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/source.wav", | |
| "bytes": 473550, | |
| "sha256": "68a5f9f7d45bb98e16e64c996fa23cc0b075d35a47f26d7a2abe56af5bc2d755", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 236736, | |
| "duration_seconds": 9.864, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/ours_45k.wav", | |
| "bytes": 952364, | |
| "sha256": "8bda349836a3d7bcc9ae608b52b616805394ef7fa60ea78bbbda9c6bd610f859", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 476160, | |
| "duration_seconds": 9.92, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/qwen3_omni.wav", | |
| "bytes": 256214, | |
| "sha256": "ccff9a3a4ce14b862feff84464c4b3a46a0460a1e06dc5270a626bec196b691d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 128085, | |
| "duration_seconds": 5.336875, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/step_audio_editx.wav", | |
| "bytes": 1025324, | |
| "sha256": "3f79b3bfe233ff1835c5e67c6533fd2f13b12f6bff198eb26bea2c014005159c", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 512640, | |
| "duration_seconds": 21.36, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/expert_data_pipeline.wav", | |
| "bytes": 497516, | |
| "sha256": "ee1a1ba86a6e5be2559badc4562b9fbd2064724601560dea4dcc864b52341751", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 248736, | |
| "duration_seconds": 10.364, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/source.wav", | |
| "bytes": 496590, | |
| "sha256": "5b57c31106fc381b3525b71625a7ec8914e26ca6dbc80db8d373674e50a934c2", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 248256, | |
| "duration_seconds": 10.344, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/ours_45k.wav", | |
| "bytes": 906284, | |
| "sha256": "4ef45ee10efce4b0db0b5872348ddac32aa3d29c9d27c54258cfa08e125fd9e7", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 453120, | |
| "duration_seconds": 9.44, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/qwen3_omni.wav", | |
| "bytes": 194774, | |
| "sha256": "5c54d49d60f661355a50caef869cb3bb6ed5916a6956534874da23eeee719446", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 97365, | |
| "duration_seconds": 4.056875, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/step_audio_editx.wav", | |
| "bytes": 420524, | |
| "sha256": "8e3719491c511ce7028901eac620f0cd9b574ee23a83ccca4bc56865030623b1", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 210240, | |
| "duration_seconds": 8.76, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/expert_data_pipeline.wav", | |
| "bytes": 455996, | |
| "sha256": "b9e30ec076f654589089f0a3ba9b80f3be94c0a4be66cb8e10cc0db021729d2d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 227976, | |
| "duration_seconds": 9.499, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/source.wav", | |
| "bytes": 764238, | |
| "sha256": "cc0fc339fcee38b63dea74a87cc877828bcf50f971ea56571e96d2b0c05de7eb", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 382080, | |
| "duration_seconds": 15.92, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/ours_45k.wav", | |
| "bytes": 1520684, | |
| "sha256": "99ddf77bc5cea6f17f371bfae6bdf89449845ab7568c7564446d41bf985f31d9", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 760320, | |
| "duration_seconds": 15.84, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/qwen3_omni.wav", | |
| "bytes": 532694, | |
| "sha256": "c75d8ec6c0d623c8e02751ed75ef1d6503723495a71eff4b7e4d857da8c4ac11", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 266325, | |
| "duration_seconds": 11.096875, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", | |
| "role": "qwen3_omni" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/step_audio_editx.wav", | |
| "bytes": 1244204, | |
| "sha256": "1868adf2a68602efe4a746c21f157429c8b3b30a41232f6faeeb4fe435ff9d8d", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 622080, | |
| "duration_seconds": 25.92, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", | |
| "role": "step_audio_editx" | |
| }, | |
| { | |
| "path": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/expert_data_pipeline.wav", | |
| "bytes": 819240, | |
| "sha256": "7cd62b65b9570123118dd4d703ca9e4e7b0ef16ea7ec2c58db715cc347fad6a7", | |
| "channels": 1, | |
| "sample_rate": 24000, | |
| "sample_width_bytes": 2, | |
| "frames": 409598, | |
| "duration_seconds": 17.066583, | |
| "benchmark": "multi", | |
| "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", | |
| "role": "expert_data_pipeline" | |
| }, | |
| { | |
| "path": "demos/noise/200638dd80c84bf1ac1c26cb873cb1c4/source.wav", | |
| "bytes": 514604, | |
| "sha256": "5fa10a59ecb8fc77a78e77b00311c4c796278e3928d6e7da72ae913026f5ffbe", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 257280, | |
| "duration_seconds": 5.36, | |
| "benchmark": "noise", | |
| "candidate_id": "200638dd80c84bf1ac1c26cb873cb1c4", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/noise/200638dd80c84bf1ac1c26cb873cb1c4/ours_45k.wav", | |
| "bytes": 522284, | |
| "sha256": "7339258ba22d0828cbd95e8c44a4e5658192c338fb8ecd913f3d75b07783f1e9", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 261120, | |
| "duration_seconds": 5.44, | |
| "benchmark": "noise", | |
| "candidate_id": "200638dd80c84bf1ac1c26cb873cb1c4", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/noise/9f20ec35b544411aab21db298ee5b6d0/source.wav", | |
| "bytes": 509228, | |
| "sha256": "05c5edc34f0b336551542e93360de105d37b839f47a1313f9381a950840377b5", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 254592, | |
| "duration_seconds": 5.304, | |
| "benchmark": "noise", | |
| "candidate_id": "9f20ec35b544411aab21db298ee5b6d0", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/noise/9f20ec35b544411aab21db298ee5b6d0/ours_45k.wav", | |
| "bytes": 506924, | |
| "sha256": "86c8b07be626d906e22acd9662a20da5295e625753403f1c7ad7fe1f7e2fc323", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 253440, | |
| "duration_seconds": 5.28, | |
| "benchmark": "noise", | |
| "candidate_id": "9f20ec35b544411aab21db298ee5b6d0", | |
| "role": "ours_45k" | |
| }, | |
| { | |
| "path": "demos/multi_speaker/e26db2858abb400f9e8ab261fcf565e9/source.wav", | |
| "bytes": 935622, | |
| "sha256": "63b9adcfafe78eca27146115d12badd86359e95570636405b6bf753326ca541d", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 467789, | |
| "duration_seconds": 9.745604, | |
| "benchmark": "multi_speaker", | |
| "candidate_id": "e26db2858abb400f9e8ab261fcf565e9", | |
| "role": "source" | |
| }, | |
| { | |
| "path": "demos/multi_speaker/e26db2858abb400f9e8ab261fcf565e9/ours_45k.wav", | |
| "bytes": 983084, | |
| "sha256": "1f2cb4faa5334e2c5e8422e4268e4189a4b510e2225bae84026a3e8a1e87d84a", | |
| "channels": 1, | |
| "sample_rate": 48000, | |
| "sample_width_bytes": 2, | |
| "frames": 491520, | |
| "duration_seconds": 10.24, | |
| "benchmark": "multi_speaker", | |
| "candidate_id": "e26db2858abb400f9e8ab261fcf565e9", | |
| "role": "ours_45k" | |
| } | |
| ] | |
| } | |