{ "schemaVersion": 1, "generatedAt": "2026-09-15T01:01:19.756Z", "purpose": "local kirin one-loop audio candidate set; not wired into EMMAAPP", "scope": "Main success loop uses Mao manual phrase-level intonation; retry branch uses existing high-rated Michinoku Airi original.", "engine": { "name": "AivisSpeech Engine", "version": "1.2.0", "host": "127.0.0.1", "port": 10101, "inference": "CPU", "useGpu": false }, "policy": { "topLevelParameters": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1 }, "topLevelChanges": "none", "globalPitchOrIntonationBoost": false, "adjustment": "Mao uses accent phrase positions and semantic sentence splits only; no global pitch/intonation boost.", "retry": "Only the retry phrase changes model; it references the human-approved Michinoku Airi original and receives edge/volume normalization for this set.", "volumeNormalization": { "thresholdRms": 0.006, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "targetRmsDbfs": -19, "peakCeilingDbfs": -1, "rebalanceTargetRmsDbfs": -20.5, "rebalancedAt": "2026-09-15T01:02:15.745Z" }, "originalFilesUntouched": true, "appIntegration": false, "noExternalApi": true }, "models": { "mao": { "id": "mao-futsu", "name": "まお", "style": "ふつー", "modelUuid": "a59cb814-0083-4369-8542-f51a29e72af7", "speakerUuid": "e756b8e4-b606-4e15-99b1-3f9c6a1b2317", "speakerId": 888753761, "version": "1.2.0", "license": "ACML 1.0", "sha256": "f87ccea2e8e2de0e0bfe52e803945af903b4086bf25621a015111628f00e4119", "page": "https://hub.aivis-project.com/aivm-models/a59cb814-0083-4369-8542-f51a29e72af7" }, "retry": { "id": "michinoku-airi-standard", "name": "みちのくあいり", "style": "標準", "modelUuid": "1b2830f4-8cf1-4184-a0d9-3a1bace3a844", "speakerUuid": "b01070b1-7243-4d5b-825f-a56ab2eb2bc2", "speakerId": 1717361472, "version": "1.0.0", "license": "ACML 1.0", "sha256": "9f329209d12b048d1da9f37b6836d47d964014906531fa206118a7e15fca03c5", "page": "https://hub.aivis-project.com/aivm-models/1b2830f4-8cf1-4184-a0d9-3a1bace3a844" } }, "sourceFiles": { "maoPriorComparison": "../model-comparison-v0.1/", "retryOriginal": "../model-comparison-v0.1/michinoku-airi-standard-04-oshii-mou-ikkai.wav", "retryOriginalSha256": "489c135dd29d41bbb39e51b636e64eab55afd3527ba714131a4a653a11968a87" }, "phrases": [ { "id": "01-kirin-wo-tsukurou", "text": "きりんを つくろう！", "slug": "kirin-wo-tsukurou" }, { "id": "02-guide-start-end", "text": "あかいまるから、ほしまで いってみよう！", "slug": "guide-start-end" }, { "id": "03-dekita", "text": "できた！", "slug": "dekita" }, { "id": "04-ii-kanji", "text": "いいかんじ！", "slug": "ii-kanji" }, { "id": "05-jouzu", "text": "じょうず！", "slug": "jouzu" }, { "id": "06-sono-choushi", "text": "そのちょうし！", "slug": "sono-choushi" }, { "id": "07-ki-dekitaa-tsugi-ri", "text": "き、できたー！つぎは、り！", "slug": "ki-dekitaa-tsugi-ri" }, { "id": "08-ri-dekita-tsugi-n", "text": "り、できた！つぎは、ん！", "slug": "ri-dekita-tsugi-n" }, { "id": "09-n-dekitaa", "text": "ん、できたー！", "slug": "n-dekitaa" }, { "id": "10-kirin-dekitaa", "text": "き・り・ん！できたー！", "slug": "kirin-dekitaa" }, { "id": "11-kimi-no-moji-kara", "text": "きみの もじから、きりんが うまれたよ！", "slug": "kimi-no-moji-kara" }, { "id": "retry-oshii", "text": "おしい！もういっかい、ここからだよ！", "slug": "oshii-mou-ikkai" } ], "profiles": { "01-kirin-wo-tsukurou": { "accents": [ 3, 2 ], "method": "accent" }, "02-guide-start-end": { "accents": [ 3, 4, 1, 4 ], "method": "accent+split", "chunks": [ "あかいまるから、", "ほしまで いってみよう！" ], "chunkAccents": [ [ 3, 4 ], [ 1, 4 ] ], "gapMs": 160 }, "03-dekita": { "accents": [ 2 ], "method": "accent" }, "04-ii-kanji": { "accents": [ 1, 2 ], "method": "accent" }, "05-jouzu": { "accents": [ 2 ], "method": "accent" }, "06-sono-choushi": { "accents": [ 3 ], "method": "accent" }, "07-ki-dekitaa-tsugi-ri": { "accents": [ 1, 2, 2, 1 ], "method": "accent+split", "chunks": [ "き、できたー！", "つぎは、り！" ], "chunkAccents": [ [ 1, 2 ], [ 2, 1 ] ], "gapMs": 130 }, "08-ri-dekita-tsugi-n": { "accents": [ 1, 2, 2, 1 ], "method": "accent+split", "chunks": [ "り、できた！", "つぎは、ん！" ], "chunkAccents": [ [ 1, 2 ], [ 2, 1 ] ], "gapMs": 130 }, "09-n-dekitaa": { "accents": [ 1, 2 ], "method": "accent" }, "10-kirin-dekitaa": { "accents": [ 1, 1, 1, 2 ], "method": "accent+split", "chunks": [ "き・り・ん！", "できたー！" ], "chunkAccents": [ [ 1, 1, 1 ], [ 2 ] ], "gapMs": 130 }, "11-kimi-no-moji-kara": { "accents": [ 5, 4, 4 ], "method": "accent+split", "chunks": [ "きみの もじから、", "きりんが うまれたよ！" ], "chunkAccents": [ [ 5 ], [ 4, 4 ] ], "gapMs": 170 } }, "sourceQueries": { "01-kirin-wo-tsukurou": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "きりんを つくろう！", "accentPhrases": [ { "text": "キリンオ", "accent": 4, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 } ] }, { "text": "ツクロオ!", "accent": 2, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ク", "vowel": "u", "pitch": 0 }, { "text": "ロ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "02-guide-start-end": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "あかいまるから、ほしまで いってみよう！", "accentPhrases": [ { "text": "アカイ", "accent": 3, "moras": [ { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "イ", "vowel": "i", "pitch": 0 } ] }, { "text": "マルカラ,", "accent": 4, "moras": [ { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "ル", "vowel": "u", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ラ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "ホシマデ", "accent": 1, "moras": [ { "text": "ホ", "vowel": "o", "pitch": 0 }, { "text": "シ", "vowel": "i", "pitch": 0 }, { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "デ", "vowel": "e", "pitch": 0 } ] }, { "text": "イッテミヨオ!", "accent": 4, "moras": [ { "text": "イ", "vowel": "i", "pitch": 0 }, { "text": "ッ", "vowel": "cl", "pitch": 0 }, { "text": "テ", "vowel": "e", "pitch": 0 }, { "text": "ミ", "vowel": "i", "pitch": 0 }, { "text": "ヨ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "03-dekita": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "できた！", "accentPhrases": [ { "text": "デキタ!", "accent": 1, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "04-ii-kanji": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "いいかんじ！", "accentPhrases": [ { "text": "イイ", "accent": 1, "moras": [ { "text": "イ", "vowel": "i", "pitch": 0 }, { "text": "イ", "vowel": "i", "pitch": 0 } ] }, { "text": "カンジ!", "accent": 1, "moras": [ { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "ジ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "05-jouzu": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "じょうず！", "accentPhrases": [ { "text": "ジョオズ!", "accent": 3, "moras": [ { "text": "ジョ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "ズ", "vowel": "u", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "06-sono-choushi": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "そのちょうし！", "accentPhrases": [ { "text": "ソノチョオシ!", "accent": 3, "moras": [ { "text": "ソ", "vowel": "o", "pitch": 0 }, { "text": "ノ", "vowel": "o", "pitch": 0 }, { "text": "チョ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "シ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "07-ki-dekitaa-tsugi-ri": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "き、できたー！つぎは、り！", "accentPhrases": [ { "text": "キ,", "accent": 2, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタア!", "accent": 1, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] }, { "text": "ツギハ,", "accent": 3, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ギ", "vowel": "i", "pitch": 0 }, { "text": "ハ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "リ!", "accent": 2, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "08-ri-dekita-tsugi-n": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "り、できた！つぎは、ん！", "accentPhrases": [ { "text": "リ,", "accent": 2, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタ!", "accent": 1, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] }, { "text": "ツギワ,", "accent": 2, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ギ", "vowel": "i", "pitch": 0 }, { "text": "ワ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "ン!", "accent": 2, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "09-n-dekitaa": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "ん、できたー！", "accentPhrases": [ { "text": "ン,", "accent": 2, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタア!", "accent": 1, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "10-kirin-dekitaa": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "き・り・ん！できたー！", "accentPhrases": [ { "text": "キ,", "accent": 2, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "リ,", "accent": 2, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "ン!", "accent": 2, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタア!", "accent": 1, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "11-kimi-no-moji-kara": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "きみの もじから、きりんが うまれたよ！", "accentPhrases": [ { "text": "キミノモジカラ,", "accent": 5, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "ミ", "vowel": "i", "pitch": 0 }, { "text": "ノ", "vowel": "o", "pitch": 0 }, { "text": "モ", "vowel": "o", "pitch": 0 }, { "text": "ジ", "vowel": "i", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ラ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "キリンガ", "accent": 4, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "ガ", "vowel": "a", "pitch": 0 } ] }, { "text": "ウマレタヨ!", "accent": 4, "moras": [ { "text": "ウ", "vowel": "u", "pitch": 0 }, { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "レ", "vowel": "e", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ヨ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } }, "tracks": [ { "sequence": 1, "branch": "main", "id": "01-kirin-wo-tsukurou", "text": "きりんを つくろう！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "01-mao-kirin-wo-tsukurou.wav", "synthesisTexts": [ "きりんを つくろう！" ], "method": "accent", "accentPattern": [ 3, 2 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "きりんを つくろう！", "accentPhrases": [ { "text": "キリンオ", "accent": 3, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 } ] }, { "text": "ツクロオ!", "accent": 2, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ク", "vowel": "u", "pitch": 0 }, { "text": "ロ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 110, "trailingTrimMs": 100, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -4.3, "rmsDbfs": -20.5, "peakDbfs": -5.8, "rebalanceGainDb": -1.5 }, "bytes": 102584, "durationSeconds": 1.163, "sha256": "6ba95bf23a711ff71c1b1a55fb5e7b18a1671118a62b75aaa9a56f0164561546" }, { "sequence": 2, "branch": "main", "id": "02-guide-start-end", "text": "あかいまるから、ほしまで いってみよう！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "02-mao-guide-start-end.wav", "synthesisTexts": [ "あかいまるから、", "ほしまで いってみよう！" ], "method": "accent+split", "accentPattern": [ 3, 4, 1, 4 ], "query": [ { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "あかいまるから、", "accentPhrases": [ { "text": "アカイ", "accent": 3, "moras": [ { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "イ", "vowel": "i", "pitch": 0 } ] }, { "text": "マルカラ,", "accent": 4, "moras": [ { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "ル", "vowel": "u", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ラ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] } ] }, { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "ほしまで いってみよう！", "accentPhrases": [ { "text": "ホシマデ", "accent": 1, "moras": [ { "text": "ホ", "vowel": "o", "pitch": 0 }, { "text": "シ", "vowel": "i", "pitch": 0 }, { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "デ", "vowel": "e", "pitch": 0 } ] }, { "text": "イッテミヨオ!", "accent": 4, "moras": [ { "text": "イ", "vowel": "i", "pitch": 0 }, { "text": "ッ", "vowel": "cl", "pitch": 0 }, { "text": "テ", "vowel": "e", "pitch": 0 }, { "text": "ミ", "vowel": "i", "pitch": 0 }, { "text": "ヨ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } ], "normalization": { "leadingTrimMs": 80, "trailingTrimMs": 90, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.98, "rmsDbfs": -20.5, "peakDbfs": -4.48, "rebalanceGainDb": -1.5 }, "bytes": 273022, "durationSeconds": 3.095, "sha256": "76b27d1405d1ff24c76066ae2cf8d286fdc0ea18e7206d4408b27e22165af67f" }, { "sequence": 3, "branch": "main", "id": "03-dekita", "text": "できた！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "03-mao-dekita.wav", "synthesisTexts": [ "できた！" ], "method": "accent", "accentPattern": [ 2 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "できた！", "accentPhrases": [ { "text": "デキタ!", "accent": 2, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 170, "trailingTrimMs": 90, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.84, "rmsDbfs": -20.5, "peakDbfs": -4.34, "rebalanceGainDb": -1.5 }, "bytes": 66426, "durationSeconds": 0.753, "sha256": "12ee36858ef297a0ca118f04da34923fc222668f89dc805d09f2664ef7fe9003" }, { "sequence": 4, "branch": "main", "id": "04-ii-kanji", "text": "いいかんじ！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "04-mao-ii-kanji.wav", "synthesisTexts": [ "いいかんじ！" ], "method": "accent", "accentPattern": [ 1, 2 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "いいかんじ！", "accentPhrases": [ { "text": "イイ", "accent": 1, "moras": [ { "text": "イ", "vowel": "i", "pitch": 0 }, { "text": "イ", "vowel": "i", "pitch": 0 } ] }, { "text": "カンジ!", "accent": 2, "moras": [ { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "ジ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 140, "trailingTrimMs": 210, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.78, "rmsDbfs": -20.5, "peakDbfs": -4.28, "rebalanceGainDb": -1.5 }, "bytes": 81016, "durationSeconds": 0.918, "sha256": "b1c8f4844cfe3ae2c0231db77dea63653c6c5861b5a89229d3d5f5c6a8be715b" }, { "sequence": 5, "branch": "main", "id": "05-jouzu", "text": "じょうず！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "05-mao-jouzu.wav", "synthesisTexts": [ "じょうず！" ], "method": "accent", "accentPattern": [ 2 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "じょうず！", "accentPhrases": [ { "text": "ジョオズ!", "accent": 2, "moras": [ { "text": "ジョ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "ズ", "vowel": "u", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 160, "trailingTrimMs": 180, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -5.91, "rmsDbfs": -20.5, "peakDbfs": -7.41, "rebalanceGainDb": -1.5 }, "bytes": 67568, "durationSeconds": 0.766, "sha256": "69a7d6638bb7ede8055a6d1db7b63434791cb5587f95f0794f1b29e2814654e7" }, { "sequence": 6, "branch": "main", "id": "06-sono-choushi", "text": "そのちょうし！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "06-mao-sono-choushi.wav", "synthesisTexts": [ "そのちょうし！" ], "method": "accent", "accentPattern": [ 3 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "そのちょうし！", "accentPhrases": [ { "text": "ソノチョオシ!", "accent": 3, "moras": [ { "text": "ソ", "vowel": "o", "pitch": 0 }, { "text": "ノ", "vowel": "o", "pitch": 0 }, { "text": "チョ", "vowel": "o", "pitch": 0 }, { "text": "オ", "vowel": "o", "pitch": 0 }, { "text": "シ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 120, "trailingTrimMs": 130, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.96, "rmsDbfs": -20.5, "peakDbfs": -4.46, "rebalanceGainDb": -1.5 }, "bytes": 100080, "durationSeconds": 1.134, "sha256": "e026deb47a106ee69828f04d47836eed36969ea553f2925a6c7077a9c2e918b8" }, { "sequence": 7, "branch": "main", "id": "07-ki-dekitaa-tsugi-ri", "text": "き、できたー！つぎは、り！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "07-mao-ki-dekitaa-tsugi-ri.wav", "synthesisTexts": [ "き、できたー！", "つぎは、り！" ], "method": "accent+split", "accentPattern": [ 1, 2, 2, 1 ], "query": [ { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "き、できたー！", "accentPhrases": [ { "text": "キ,", "accent": 1, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタア!", "accent": 2, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "つぎは、り！", "accentPhrases": [ { "text": "ツギハ,", "accent": 2, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ギ", "vowel": "i", "pitch": 0 }, { "text": "ハ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "リ!", "accent": 1, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } ], "normalization": { "leadingTrimMs": 110, "trailingTrimMs": 110, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.3, "rmsDbfs": -20.5, "peakDbfs": -3.8, "rebalanceGainDb": -1.5 }, "bytes": 238310, "durationSeconds": 2.701, "sha256": "8d2d7618b842875d4a5b1ddbd858af97952132de5f74acd8d1576e49eba6573d" }, { "sequence": 8, "branch": "main", "id": "08-ri-dekita-tsugi-n", "text": "り、できた！つぎは、ん！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "08-mao-ri-dekita-tsugi-n.wav", "synthesisTexts": [ "り、できた！", "つぎは、ん！" ], "method": "accent+split", "accentPattern": [ 1, 2, 2, 1 ], "query": [ { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "り、できた！", "accentPhrases": [ { "text": "リ,", "accent": 1, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタ!", "accent": 2, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "つぎは、ん！", "accentPhrases": [ { "text": "ツギワ,", "accent": 2, "moras": [ { "text": "ツ", "vowel": "u", "pitch": 0 }, { "text": "ギ", "vowel": "i", "pitch": 0 }, { "text": "ワ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "ン!", "accent": 1, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } ], "normalization": { "leadingTrimMs": 110, "trailingTrimMs": 130, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -1.86, "rmsDbfs": -20.5, "peakDbfs": -3.36, "rebalanceGainDb": -1.5 }, "bytes": 207892, "durationSeconds": 2.357, "sha256": "a8a281214eec076643cb1d62dafe8a1d1ac111ead40fe4881fd2fb1c32dd1109" }, { "sequence": 9, "branch": "main", "id": "09-n-dekitaa", "text": "ん、できたー！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "09-mao-n-dekitaa.wav", "synthesisTexts": [ "ん、できたー！" ], "method": "accent", "accentPattern": [ 1, 2 ], "query": { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "ん、できたー！", "accentPhrases": [ { "text": "ン,", "accent": 1, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "デキタア!", "accent": 2, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, "normalization": { "leadingTrimMs": 120, "trailingTrimMs": 100, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -1.97, "rmsDbfs": -20.5, "peakDbfs": -3.47, "rebalanceGainDb": -1.5 }, "bytes": 106918, "durationSeconds": 1.212, "sha256": "7bbe128ba669a64491d9c7d67641cdab72059a8b003ad26245b033bfe1eabf33" }, { "sequence": 10, "branch": "main", "id": "10-kirin-dekitaa", "text": "き・り・ん！できたー！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "10-mao-kirin-dekitaa.wav", "synthesisTexts": [ "き・り・ん！", "できたー！" ], "method": "accent+split", "accentPattern": [ 1, 1, 1, 2 ], "query": [ { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "き・り・ん！", "accentPhrases": [ { "text": "キ,", "accent": 1, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "リ,", "accent": 1, "moras": [ { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] }, { "text": "ン!", "accent": 1, "moras": [ { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] }, { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "できたー！", "accentPhrases": [ { "text": "デキタア!", "accent": 2, "moras": [ { "text": "デ", "vowel": "e", "pitch": 0 }, { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ア", "vowel": "a", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } ], "normalization": { "leadingTrimMs": 50, "trailingTrimMs": 90, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -2.56, "rmsDbfs": -20.5, "peakDbfs": -4.06, "rebalanceGainDb": -1.5 }, "bytes": 212604, "durationSeconds": 2.41, "sha256": "48dd0ce4eeaea8845d1b34425c29cb14893f875d712166355a89bee52434a81e" }, { "sequence": 11, "branch": "main", "id": "11-kimi-no-moji-kara", "text": "きみの もじから、きりんが うまれたよ！", "modelId": "mao-futsu", "model": "まお", "style": "ふつー", "file": "11-mao-kimi-no-moji-kara.wav", "synthesisTexts": [ "きみの もじから、", "きりんが うまれたよ！" ], "method": "accent+split", "accentPattern": [ 5, 4, 4 ], "query": [ { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "きみの もじから、", "accentPhrases": [ { "text": "キミノモジカラ,", "accent": 5, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "ミ", "vowel": "i", "pitch": 0 }, { "text": "ノ", "vowel": "o", "pitch": 0 }, { "text": "モ", "vowel": "o", "pitch": 0 }, { "text": "ジ", "vowel": "i", "pitch": 0 }, { "text": "カ", "vowel": "a", "pitch": 0 }, { "text": "ラ", "vowel": "a", "pitch": 0 }, { "text": ",", "vowel": "pau", "pitch": 0 } ] } ] }, { "speedScale": 1, "pitchScale": 0, "intonationScale": 1, "tempoDynamicsScale": 1, "volumeScale": 1, "prePhonemeLength": 0.1, "postPhonemeLength": 0.1, "pauseLength": null, "pauseLengthScale": 1, "kana": "きりんが うまれたよ！", "accentPhrases": [ { "text": "キリンガ", "accent": 4, "moras": [ { "text": "キ", "vowel": "i", "pitch": 0 }, { "text": "リ", "vowel": "i", "pitch": 0 }, { "text": "ン", "vowel": "N", "pitch": 0 }, { "text": "ガ", "vowel": "a", "pitch": 0 } ] }, { "text": "ウマレタヨ!", "accent": 4, "moras": [ { "text": "ウ", "vowel": "u", "pitch": 0 }, { "text": "マ", "vowel": "a", "pitch": 0 }, { "text": "レ", "vowel": "e", "pitch": 0 }, { "text": "タ", "vowel": "a", "pitch": 0 }, { "text": "ヨ", "vowel": "o", "pitch": 0 }, { "text": "!", "vowel": "pau", "pitch": 0 } ] } ] } ], "normalization": { "leadingTrimMs": 110, "trailingTrimMs": 160, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -3.61, "rmsDbfs": -20.5, "peakDbfs": -5.11, "rebalanceGainDb": -1.5 }, "bytes": 277382, "durationSeconds": 3.144, "sha256": "122715b2fda3c917144ba7809d423428d0ef6f2f0edb9e6907b73e521edc45ce" }, { "sequence": 12, "branch": "retry", "id": "retry-oshii", "text": "おしい！もういっかい、ここからだよ！", "modelId": "michinoku-airi-standard", "model": "みちのくあいり", "style": "標準", "file": "12-michinoku-airi-oshii-mou-ikkai.wav", "sourceFile": "../model-comparison-v0.1/michinoku-airi-standard-04-oshii-mou-ikkai.wav", "method": "existing high-rated original + volume/edge normalization only", "query": null, "normalization": { "leadingTrimMs": 30, "trailingTrimMs": 0, "retainedLeadingPadMs": 80, "retainedTrailingPadMs": 120, "gainDb": -1, "rmsDbfs": -20.81, "peakDbfs": -1, "rebalanceGainDb": 0 }, "bytes": 319166, "durationSeconds": 3.618, "sha256": "4a273f65c4c8a70c20e0fa6bf69760b8190cd8266369542445c68f864df2e0f7" } ], "generatedFiles": [ "01-mao-kirin-wo-tsukurou.wav", "02-mao-guide-start-end.wav", "03-mao-dekita.wav", "04-mao-ii-kanji.wav", "05-mao-jouzu.wav", "06-mao-sono-choushi.wav", "07-mao-ki-dekitaa-tsugi-ri.wav", "08-mao-ri-dekita-tsugi-n.wav", "09-mao-n-dekitaa.wav", "10-mao-kirin-dekitaa.wav", "11-mao-kimi-no-moji-kara.wav", "12-michinoku-airi-oshii-mou-ikkai.wav" ], "notes": [ "The main sequence is a successful one-play path. Retry is a branch and is placed last for side-by-side continuity testing.", "Kohaku is not regenerated here; this set follows the human-approved Mao adjustment and preserves existing Kohaku materials separately.", "AudioQuery mora pitch values are not used for the main set; the proven accent/split method is the primary path." ] }