{ "schema_version": 1, "title": "dots.tts.edit", "catalog_fingerprint": "746b255476160936117a35a21dc2ded1b803bb26bf6046e16ed87c98281eb8ff", "selection_sha256": "ac2de6a310092df70ba262743459ca7e2f837db1ab263d793f5a3746e88da5d5", "sample_count": 29, "audio_count": 136, "sections": [ { "benchmark": "text", "title": "Text Editing", "samples": [ { "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0032", "canonical_id": "text_edit__easy-0727__0727_en_0032", "language": "en", "focus": "Two localized substitutions while preserving the speaker's distinctive pronunciation and surrounding prosody.", "tags": [ "en", "2 ops", "bulk_sub" ], "instruction_html": "
"How I wishReplace with “long” you could getReplace with “obtain” me a copy of that picture, Philip," Laura said, entreatingly.
", "source_audio": "demos/text/text_edit__easy-0727__0727_en_0032/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__easy-0727__0727_en_0032/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0032/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0032/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0032/expert_data_pipeline.wav" } ] }, { "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_en_0057", "canonical_id": "text_edit__easy-0727__0727_en_0057", "language": "en", "focus": "A single long-span edit at the end of an extended English utterance.", "tags": [ "en", "1 ops", "sub", "long_source" ], "instruction_html": "
The exigencies of the modern industrial system frequently place individuals and households in juxtaposition between whom there is little contact in any other sense than that of juxtaposition. One's neighbors, mechanically speaking, often are socially not one's neighbors, or even acquaintances; and still their transient good opinion has a high degree of utilityReplace with “brief approval remains highly useful”.
", "source_audio": "demos/text/text_edit__easy-0727__0727_en_0057/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__easy-0727__0727_en_0057/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0057/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0057/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_en_0057/expert_data_pipeline.wav" } ] }, { "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0052", "canonical_id": "text_edit__easy-0727__0727_zh_0052", "language": "zh", "focus": "Mandarin substitution plus insertion, with the original mouse click and pauses left intact.", "tags": [ "zh", "2 ops", "mixed", "insert_present", "expansion" ], "instruction_html": "
本届亚运会上没有获得任何奖牌Replace with “金牌”Insert “但多名选手进入了决赛。”
", "source_audio": "demos/text/text_edit__easy-0727__0727_zh_0052/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__easy-0727__0727_zh_0052/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0052/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0052/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0052/expert_data_pipeline.wav" } ] }, { "candidate_id": "text:easy-0727:text_edit__easy-0727__0727_zh_0075", "canonical_id": "text_edit__easy-0727__0727_zh_0075", "language": "zh", "focus": "A localized substitution inside a long Mandarin list.", "tags": [ "zh", "1 ops", "bulk_sub" ], "instruction_html": "
其他提名包括最佳影片、最佳导演、最佳摄影、最佳服装设计、最佳电影剪辑、最佳原创配乐、最佳艺术指导、最佳音效剪辑、最佳混音和最佳原创剧本Replace with “最佳改编剧本”
", "source_audio": "demos/text/text_edit__easy-0727__0727_zh_0075/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__easy-0727__0727_zh_0075/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0075/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0075/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__easy-0727__0727_zh_0075/expert_data_pipeline.wav" } ] }, { "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_en_0107", "canonical_id": "text_edit__hard-0727__hard_0727_en_0107", "language": "en", "focus": "Three operations rewrite nearly the entire sentence while retaining the source speaker style.", "tags": [ "en", "3 ops", "mixed", "acronym_or_allcaps", "multiop_ge3", "delete_present" ], "instruction_html": "
For, it must be stated to those who are ignorant of the caseReplace with “NASA ”Insert “changed”, that argot is both a literary phenomenon and a social resultDelete.
", "source_audio": "demos/text/text_edit__hard-0727__hard_0727_en_0107/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__hard-0727__hard_0727_en_0107/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_en_0107/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_en_0107/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_en_0107/expert_data_pipeline.wav" } ] }, { "candidate_id": "text:hard-0727:text_edit__hard-0727__hard_0727_zh_0208", "canonical_id": "text_edit__hard-0727__hard_0727_zh_0208", "language": "zh", "focus": "Two long-span Mandarin substitutions compress the source into a short target.", "tags": [ "zh", "2 ops", "mixed", "chinese_numerals", "compression" ], "instruction_html": "
摄政王广场由摄政王大街呈东西向贯通Replace with “结果”,另有南北六条较小的街道与广场相通Replace with “变了”
", "source_audio": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/qwen3_omni.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/mimo_audio_instruct.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/text/text_edit__hard-0727__hard_0727_zh_0208/expert_data_pipeline.wav" } ] } ] }, { "benchmark": "emotion", "title": "Emotion Editing", "samples": [ { "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_043_phrase", "canonical_id": "emotion_edit__curated-312__emo_c312_043_phrase", "language": "en", "focus": "Localized anger on one English phrase while the surrounding context remains neutral.", "tags": [ "en", "single_phrase", "angry" ], "instruction_html": "
When God made these different races he knew that the superior would enslave the inferior, knew that the inferior would be conqueredEmotion · Angry · level 2, and finally destroyed.
", "source_audio": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/ours_45k.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/step_audio_editx.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/ming_uniaudio.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_043_phrase/expert_data_pipeline.wav" } ] }, { "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_012_phrase", "canonical_id": "emotion_edit__curated-312__emo_c312_012_phrase", "language": "en", "focus": "A natural phrase-level sadness edit aligned with the sentence semantics.", "tags": [ "en", "single_phrase", "sad" ], "instruction_html": "
"Mary, your father and I are going out to ride this afternoon, and I am sorry, for your sakeEmotion · Sad · level 2, that we can not take you with us."
", "source_audio": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/ours_45k.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/step_audio_editx.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/ming_uniaudio.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_012_phrase/expert_data_pipeline.wav" } ] }, { "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_057_phrase", "canonical_id": "emotion_edit__curated-312__emo_c312_057_phrase", "language": "zh", "focus": "Localized low-arousal sadness in Mandarin.", "tags": [ "zh", "single_phrase", "sad" ], "instruction_html": "
然而列宁因为他的妹妹奥利加因伤寒逝世Emotion · Sad · level 2而未能出席毕业典礼。
", "source_audio": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/ours_45k.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/step_audio_editx.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/ming_uniaudio.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_057_phrase/expert_data_pipeline.wav" } ] }, { "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_080_multi", "canonical_id": "emotion_edit__curated-312__emo_c312_080_multi", "language": "zh", "focus": "Two Mandarin spans receive different emotions: anger and fear.", "tags": [ "zh", "multi_span", "angry+afraid" ], "instruction_html": "
还包括禁毒行动和处理Emotion · Angry · level 2大规模毁灭性武器的恐怖活动所造成Emotion · Afraid · level 2的后果。
", "source_audio": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/ours_45k.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/step_audio_editx.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/ming_uniaudio.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_080_multi/expert_data_pipeline.wav" } ] }, { "candidate_id": "emotion:curated-312:emotion_edit__curated-312__emo_c312_010_whole", "canonical_id": "emotion_edit__curated-312__emo_c312_010_whole", "language": "en", "focus": "Whole-sentence melancholic rendering for contrast with localized emotion edits.", "tags": [ "en", "whole_sentence", "melancholic" ], "instruction_html": "
To be or not to be, that is the question; whether 'tis nobler in the mind to suffer the slings and arrows-What? No, Hamlet speaking.Emotion · Melancholic · level 2
", "source_audio": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/ours_45k.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/step_audio_editx.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/ming_uniaudio.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/emotion/emotion_edit__curated-312__emo_c312_010_whole/expert_data_pipeline.wav" } ] } ] }, { "benchmark": "prosody", "title": "Prosody Editing", "samples": [ { "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0054_wave_2", "canonical_id": "prosody_edit__medium__en_input_0054_wave_2", "language": "en", "focus": "Three English regions combine slower rate, higher pitch, and faster rate.", "tags": [ "en", "mixed", "clause", "3 regions" ], "instruction_html": "
But, in spite of my blindness,Rate · 0.81× I will be caughtPitch · +4 semitones neither in her net nor in the Queen's.Rate · 1.2×
", "source_audio": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/ours_45k.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/ming_uniaudio.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0054_wave_2/expert_data_pipeline.wav" } ] }, { "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0038_wave_2", "canonical_id": "prosody_edit__medium__en_input_0038_wave_2", "language": "en", "focus": "Three pitch regions move in different directions within one sentence.", "tags": [ "en", "pitch", "clause", "3 regions" ], "instruction_html": "
As our boat flashed down the rollersPitch · +5 semitones into the waterPitch · -4 semitones I made this songPitch · +3 semitones and sang it:
", "source_audio": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/ours_45k.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/ming_uniaudio.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0038_wave_2/expert_data_pipeline.wav" } ] }, { "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0035_wave_3", "canonical_id": "prosody_edit__medium__zh_input_0035_wave_3", "language": "zh", "focus": "Two Mandarin pitch regions shift emphasis to different words.", "tags": [ "zh", "pitch", "clause", "2 regions" ], "instruction_html": "
于是我决定就来谈谈Pitch · -5 semitones种族问题Pitch · +9 semitones
", "source_audio": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/ours_45k.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/ming_uniaudio.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0035_wave_3/expert_data_pipeline.wav" } ] }, { "candidate_id": "prosody:medium:prosody_edit__medium__en_input_0019_wave_2", "canonical_id": "prosody_edit__medium__en_input_0019_wave_2", "language": "en", "focus": "A single region controls the speaking rate of an entire long English sentence.", "tags": [ "en", "rate", "sentence", "1 regions" ], "instruction_html": "
This service Plautus rendered, consciously or unconsciously, by making two Carthaginian soldiers talk Phoenician; that service Moliere rendered, by making so many of his characters talk Levantine and all sorts of dialects.Rate · 0.88×
", "source_audio": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/ours_45k.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/ming_uniaudio.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__en_input_0019_wave_2/expert_data_pipeline.wav" } ] }, { "candidate_id": "prosody:medium:prosody_edit__medium__zh_input_0019_wave_3", "canonical_id": "prosody_edit__medium__zh_input_0019_wave_3", "language": "zh", "focus": "Whole-utterance rate control over a Mandarin digit sequence.", "tags": [ "zh", "rate", "sentence", "1 regions" ], "instruction_html": "
九四零七五一幺九七三九Rate · 1.19×
", "source_audio": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/ours_45k.wav" }, { "system_id": "ming_uniaudio", "label": "Ming-UniAudio", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/ming_uniaudio.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/prosody/prosody_edit__medium__zh_input_0019_wave_3/expert_data_pipeline.wav" } ] } ] }, { "benchmark": "pause", "title": "Pause Editing", "samples": [ { "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0001", "canonical_id": "pause_edit_benchmark__full-0727__pe_0001", "language": "en", "focus": "Insert one long pause at a natural English boundary.", "tags": [ "en", "ins", "1 edits", "levels=3" ], "instruction_html": "
He threw it down at the approach of mrs Luna, laughed, Insert pause · level 3shook hands with her, and said in answer to her last remark, "You imply that you do tell fibs.
", "source_audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/ours_45k.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/mimo_audio_instruct.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0001/expert_data_pipeline.wav" } ] }, { "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0158", "canonical_id": "pause_edit_benchmark__full-0727__pe_0158", "language": "zh", "focus": "Insert one long pause at a Mandarin phrase boundary.", "tags": [ "zh", "ins", "1 edits", "levels=3" ], "instruction_html": "
但是非常考虑到整体的一个Insert pause · level 3家用的
", "source_audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/ours_45k.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/mimo_audio_instruct.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0158/expert_data_pipeline.wav" } ] }, { "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0190", "canonical_id": "pause_edit_benchmark__full-0727__pe_0190", "language": "zh", "focus": "Insert two short pauses within one Mandarin utterance.", "tags": [ "zh", "ins", "2 edits", "levels=1,1" ], "instruction_html": "
Insert pause · level 1消费者得以全面地了解和体验李宁Insert pause · level 1专业运动产品
", "source_audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/ours_45k.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/mimo_audio_instruct.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0190/expert_data_pipeline.wav" } ] }, { "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0083", "canonical_id": "pause_edit_benchmark__full-0727__pe_0083", "language": "en", "focus": "Reduce one existing pause in English.", "tags": [ "en", "red", "1 edits", "levels=2" ], "instruction_html": "
I shouldn't wonder- Reduce pause · level 2But they've probably spoiled her, so that she'd be tiresome if one met her again."
", "source_audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/ours_45k.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/mimo_audio_instruct.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0083/expert_data_pipeline.wav" } ] }, { "candidate_id": "pause:full-0727:pause_edit_benchmark__full-0727__pe_0135", "canonical_id": "pause_edit_benchmark__full-0727__pe_0135", "language": "en", "focus": "Reduce three English pauses with fine-grained timing control.", "tags": [ "en", "red", "3 edits", "levels=1,1,1" ], "instruction_html": "
These rapscallions wanted to try the Nonesuch again, Reduce pause · level 1because there was so much money in it, Reduce pause · level 1but they judged it wouldn't be safe, Reduce pause · level 1because maybe the news might a worked along down by this time.
", "source_audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/ours_45k.wav" }, { "system_id": "mimo_audio_instruct", "label": "MiMo Audio Instruct", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/mimo_audio_instruct.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/qwen3_omni.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/pause/pause_edit_benchmark__full-0727__pe_0135/expert_data_pipeline.wav" } ] } ] }, { "benchmark": "multi", "title": "Compositional Editing", "samples": [ { "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0050", "canonical_id": "multi_edit__balanced-240__multi_edit_balanced-240_me_0050", "language": "en", "focus": "Combine English pause reduction with localized rate control.", "tags": [ "en", "prosody+pause", "2 components" ], "instruction_html": "
But, Reduce pause · level 2I thank God, an imagination took him one morning that he was falling into a dropsyRate · 1.25×, and made him in such haste to go back to Cambridge to his doctor, that he never remembers anything he has to ask of me, but the coach to carry him away.
", "source_audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/qwen3_omni.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/step_audio_editx.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0050/expert_data_pipeline.wav" } ] }, { "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0004", "canonical_id": "multi_edit__balanced-240__multi_edit_balanced-240_me_0004", "language": "en", "focus": "Combine an English pause edit with localized melancholic emotion.", "tags": [ "en", "emotion+pause", "2 components" ], "instruction_html": "
So great was the cry of the wounded, that the prior of the abbey with all his monks came forth, Reduce pause · level 2who, when they saw these poor wretches so slain amongst the vinesEmotion · Melancholic · level 2, and wounded to death, confessed some of them.
", "source_audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/qwen3_omni.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/step_audio_editx.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0004/expert_data_pipeline.wav" } ] }, { "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0165", "canonical_id": "multi_edit__balanced-240__multi_edit_balanced-240_me_0165", "language": "zh", "focus": "Combine Mandarin pitch control with pause insertion.", "tags": [ "zh", "prosody+pause", "2 components" ], "instruction_html": "
让市民自行选择非制订排列Pitch · -4 semitones英文字母Insert pause · level 2及阿拉伯数字作为车辆登记号码。
", "source_audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/qwen3_omni.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/step_audio_editx.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0165/expert_data_pipeline.wav" } ] }, { "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0214", "canonical_id": "multi_edit__balanced-240__multi_edit_balanced-240_me_0214", "language": "zh", "focus": "Combine Mandarin text substitution with pause insertion.", "tags": [ "zh", "text+pause", "2 components" ], "instruction_html": "
法团Insert pause · level 2亦同时向时任特首的梁振英Replace with “向相关部门”及城市大学动物医学院的捐助者求助。
", "source_audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/qwen3_omni.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/step_audio_editx.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0214/expert_data_pipeline.wav" } ] }, { "candidate_id": "multi:balanced-240:multi_edit__balanced-240__multi_edit_balanced-240_me_0198", "canonical_id": "multi_edit__balanced-240__multi_edit_balanced-240_me_0198", "language": "zh", "focus": "Three-way Mandarin edit combining text, emotion, and speaking rate.", "tags": [ "zh", "text+emotion+prosody", "3 components" ], "instruction_html": "
就在墨线上拴上一根银针插在脚下Rate · 0.8×在变换方位后继续牵引墨线Emotion · Calm · level 2一共走了九九八十一步九根银针Replace with “三枚铜钱”被我以品字形全部插好
", "source_audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Our model · 45k", "kind": "ours", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/ours_45k.wav" }, { "system_id": "qwen3_omni", "label": "Qwen3-Omni", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/qwen3_omni.wav" }, { "system_id": "step_audio_editx", "label": "Step-Audio-EditX", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/step_audio_editx.wav" }, { "system_id": "expert_data_pipeline", "label": "Expert Data Pipeline", "kind": "baseline", "audio": "demos/multi/multi_edit__balanced-240__multi_edit_balanced-240_me_0198/expert_data_pipeline.wav" } ] } ] }, { "benchmark": "noise", "title": "Noise Preservation & Denoising", "samples": [ { "candidate_id": "200638dd80c84bf1ac1c26cb873cb1c4", "canonical_id": "200638dd80c84bf1ac1c26cb873cb1c4", "language": "en", "focus": "Text editing while preserving continuous background noise.", "tags": [ "English", "Noise preservation" ], "instruction_html": "
"How I wishReplace with “hope” you could get me a copy of that picture, Philip," Laura said, entreatingly.
", "source_audio": "demos/noise/200638dd80c84bf1ac1c26cb873cb1c4/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Ours · 45k", "kind": "ours", "audio": "demos/noise/200638dd80c84bf1ac1c26cb873cb1c4/ours_45k.wav" } ] }, { "candidate_id": "9f20ec35b544411aab21db298ee5b6d0", "canonical_id": "9f20ec35b544411aab21db298ee5b6d0", "language": "zh", "focus": "Denoising while preserving the original content and speaker identity.", "tags": [ "Mandarin", "Denoising" ], "instruction_html": "
本届亚运会上没有获得任何奖牌。Enhance
", "source_audio": "demos/noise/9f20ec35b544411aab21db298ee5b6d0/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Ours · 45k", "kind": "ours", "audio": "demos/noise/9f20ec35b544411aab21db298ee5b6d0/ours_45k.wav" } ] } ] }, { "benchmark": "multi_speaker", "title": "Multi-speaker Preservation", "samples": [ { "candidate_id": "e26db2858abb400f9e8ab261fcf565e9", "canonical_id": "e26db2858abb400f9e8ab261fcf565e9", "language": "zh", "focus": "Editing one speaker while preserving the second speaker.", "tags": [ "Mandarin", "Two speakers" ], "instruction_html": "
好像看见了一个红头发Replace with “蓝头发”的家伙。 上次你说鞋子有点磨脚,我给你买了一双软软的鞋垫,这样你垫上之后应该会舒服很多。
", "source_audio": "demos/multi_speaker/e26db2858abb400f9e8ab261fcf565e9/source.wav", "systems": [ { "system_id": "ours_45k", "label": "Ours · 45k", "kind": "ours", "audio": "demos/multi_speaker/e26db2858abb400f9e8ab261fcf565e9/ours_45k.wav" } ] } ] } ] }