@kongyo2/kongyoroid 1.0.0 → 2.0.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +233 -0
- package/dist/cache.d.ts +32 -7
- package/dist/cache.d.ts.map +1 -1
- package/dist/cache.js +138 -9
- package/dist/cache.js.map +1 -1
- package/dist/cli/flags.d.ts +270 -0
- package/dist/cli/flags.d.ts.map +1 -0
- package/dist/cli/flags.js +240 -0
- package/dist/cli/flags.js.map +1 -0
- package/dist/cli/help.d.ts +11 -0
- package/dist/cli/help.d.ts.map +1 -0
- package/dist/cli/help.js +336 -0
- package/dist/cli/help.js.map +1 -0
- package/dist/cli.js +598 -232
- package/dist/cli.js.map +1 -1
- package/dist/errors.d.ts +43 -3
- package/dist/errors.d.ts.map +1 -1
- package/dist/errors.js +56 -3
- package/dist/errors.js.map +1 -1
- package/dist/index.d.ts +59 -22
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +32 -14
- package/dist/index.js.map +1 -1
- package/dist/io.d.ts +11 -0
- package/dist/io.d.ts.map +1 -1
- package/dist/io.js +149 -5
- package/dist/io.js.map +1 -1
- package/dist/kongyoroid.d.ts +237 -10
- package/dist/kongyoroid.d.ts.map +1 -1
- package/dist/kongyoroid.js +777 -86
- package/dist/kongyoroid.js.map +1 -1
- package/dist/limits.d.ts +5 -0
- package/dist/limits.d.ts.map +1 -1
- package/dist/limits.js +6 -1
- package/dist/limits.js.map +1 -1
- package/dist/request.d.ts +1 -0
- package/dist/request.d.ts.map +1 -1
- package/dist/request.js +102 -15
- package/dist/request.js.map +1 -1
- package/dist/schema.d.ts +3 -0
- package/dist/schema.d.ts.map +1 -1
- package/dist/schema.js +271 -68
- package/dist/schema.js.map +1 -1
- package/dist/song/mml.d.ts +15 -0
- package/dist/song/mml.d.ts.map +1 -0
- package/dist/song/mml.js +184 -0
- package/dist/song/mml.js.map +1 -0
- package/dist/song/pitch.d.ts.map +1 -0
- package/dist/{pitch.js → song/pitch.js} +1 -1
- package/dist/song/pitch.js.map +1 -0
- package/dist/song/score.d.ts +92 -0
- package/dist/song/score.d.ts.map +1 -0
- package/dist/song/score.js +454 -0
- package/dist/song/score.js.map +1 -0
- package/dist/synth/acoustics.d.ts +27 -0
- package/dist/synth/acoustics.d.ts.map +1 -0
- package/dist/synth/acoustics.js +608 -0
- package/dist/synth/acoustics.js.map +1 -0
- package/dist/synth/analysis.d.ts +22 -0
- package/dist/synth/analysis.d.ts.map +1 -0
- package/dist/synth/analysis.js +170 -0
- package/dist/synth/analysis.js.map +1 -0
- package/dist/synth/engine.d.ts +115 -0
- package/dist/synth/engine.d.ts.map +1 -0
- package/dist/synth/engine.js +223 -0
- package/dist/synth/engine.js.map +1 -0
- package/dist/synth/filters.d.ts +60 -0
- package/dist/synth/filters.d.ts.map +1 -0
- package/dist/synth/filters.js +166 -0
- package/dist/synth/filters.js.map +1 -0
- package/dist/synth/glottal.d.ts +20 -0
- package/dist/synth/glottal.d.ts.map +1 -0
- package/dist/synth/glottal.js +76 -0
- package/dist/synth/glottal.js.map +1 -0
- package/dist/synth/phonemes.d.ts +42 -0
- package/dist/synth/phonemes.d.ts.map +1 -0
- package/dist/synth/phonemes.js +311 -0
- package/dist/synth/phonemes.js.map +1 -0
- package/dist/synth/plan.d.ts +127 -0
- package/dist/synth/plan.d.ts.map +1 -0
- package/dist/synth/plan.js +81 -0
- package/dist/synth/plan.js.map +1 -0
- package/dist/synth/prosody.d.ts +30 -0
- package/dist/synth/prosody.d.ts.map +1 -0
- package/dist/synth/prosody.js +46 -0
- package/dist/synth/prosody.js.map +1 -0
- package/dist/synth/random.d.ts +15 -0
- package/dist/synth/random.d.ts.map +1 -0
- package/dist/synth/random.js +59 -0
- package/dist/synth/random.js.map +1 -0
- package/dist/synth/renderer.d.ts +77 -0
- package/dist/synth/renderer.d.ts.map +1 -0
- package/dist/synth/renderer.js +343 -0
- package/dist/synth/renderer.js.map +1 -0
- package/dist/synth/song.d.ts +30 -0
- package/dist/synth/song.d.ts.map +1 -0
- package/dist/synth/song.js +453 -0
- package/dist/synth/song.js.map +1 -0
- package/dist/synth/speech.d.ts +49 -0
- package/dist/synth/speech.d.ts.map +1 -0
- package/dist/synth/speech.js +354 -0
- package/dist/synth/speech.js.map +1 -0
- package/dist/synth/stream.d.ts +26 -0
- package/dist/synth/stream.d.ts.map +1 -0
- package/dist/synth/stream.js +67 -0
- package/dist/synth/stream.js.map +1 -0
- package/dist/synth/textstream.d.ts +35 -0
- package/dist/synth/textstream.d.ts.map +1 -0
- package/dist/synth/textstream.js +127 -0
- package/dist/synth/textstream.js.map +1 -0
- package/dist/synth/voice.d.ts +64 -0
- package/dist/synth/voice.d.ts.map +1 -0
- package/dist/synth/voice.js +208 -0
- package/dist/synth/voice.js.map +1 -0
- package/dist/text/dictionary.d.ts +50 -0
- package/dist/text/dictionary.d.ts.map +1 -0
- package/dist/text/dictionary.js +0 -0
- package/dist/text/dictionary.js.map +1 -0
- package/dist/text/frontend.d.ts +34 -0
- package/dist/text/frontend.d.ts.map +1 -0
- package/dist/text/frontend.js +550 -0
- package/dist/text/frontend.js.map +1 -0
- package/dist/text/lexicon.d.ts +10 -0
- package/dist/text/lexicon.d.ts.map +1 -0
- package/dist/text/lexicon.js +763 -0
- package/dist/text/lexicon.js.map +1 -0
- package/dist/{mora.d.ts → text/mora.d.ts} +15 -3
- package/dist/text/mora.d.ts.map +1 -0
- package/dist/{mora.js → text/mora.js} +87 -9
- package/dist/text/mora.js.map +1 -0
- package/dist/text/normalize.d.ts +34 -0
- package/dist/text/normalize.d.ts.map +1 -0
- package/dist/text/normalize.js +496 -0
- package/dist/text/normalize.js.map +1 -0
- package/dist/text/notation.d.ts +32 -0
- package/dist/text/notation.d.ts.map +1 -0
- package/dist/text/notation.js +182 -0
- package/dist/text/notation.js.map +1 -0
- package/dist/text/reading.d.ts +52 -0
- package/dist/text/reading.d.ts.map +1 -0
- package/dist/text/reading.js +32 -0
- package/dist/text/reading.js.map +1 -0
- package/dist/text/sentences.d.ts +10 -0
- package/dist/text/sentences.d.ts.map +1 -0
- package/dist/text/sentences.js +84 -0
- package/dist/text/sentences.js.map +1 -0
- package/dist/{text.d.ts → text/split.d.ts} +2 -2
- package/dist/text/split.d.ts.map +1 -0
- package/dist/{text.js → text/split.js} +2 -2
- package/dist/text/split.js.map +1 -0
- package/dist/types.d.ts +58 -3
- package/dist/types.d.ts.map +1 -1
- package/dist/version.d.ts +4 -0
- package/dist/version.d.ts.map +1 -1
- package/dist/version.js +5 -1
- package/dist/version.js.map +1 -1
- package/dist/voicevox/client.d.ts +1 -1
- package/dist/voicevox/client.d.ts.map +1 -1
- package/dist/voicevox/dictionary.js +1 -1
- package/dist/voicevox/dictionary.js.map +1 -1
- package/dist/voicevox/song.js +1 -1
- package/dist/voicevox/song.js.map +1 -1
- package/dist/voicevox/speech.d.ts +1 -0
- package/dist/voicevox/speech.d.ts.map +1 -1
- package/dist/voicevox/speech.js +8 -2
- package/dist/voicevox/speech.js.map +1 -1
- package/docs/cli.md +251 -0
- package/examples/batch.jsonl +4 -3
- package/examples/builtin/dictionary.json +7 -0
- package/examples/builtin/song-compact.json +13 -0
- package/examples/builtin/song-mml.json +9 -0
- package/examples/builtin/song-notes.json +25 -0
- package/examples/builtin/speech-dictionary.json +9 -0
- package/examples/builtin/speech-kana.json +7 -0
- package/examples/builtin/speech.json +7 -0
- package/package.json +11 -3
- package/dist/formant.d.ts +0 -34
- package/dist/formant.d.ts.map +0 -1
- package/dist/formant.js +0 -622
- package/dist/formant.js.map +0 -1
- package/dist/mora.d.ts.map +0 -1
- package/dist/mora.js.map +0 -1
- package/dist/notation.d.ts +0 -15
- package/dist/notation.d.ts.map +0 -1
- package/dist/notation.js +0 -114
- package/dist/notation.js.map +0 -1
- package/dist/pitch.d.ts.map +0 -1
- package/dist/pitch.js.map +0 -1
- package/dist/score.d.ts +0 -37
- package/dist/score.d.ts.map +0 -1
- package/dist/score.js +0 -166
- package/dist/score.js.map +0 -1
- package/dist/text.d.ts.map +0 -1
- package/dist/text.js.map +0 -1
- package/examples/song-notes.json +0 -24
- /package/dist/{pitch.d.ts → song/pitch.d.ts} +0 -0
- /package/examples/{song.json → voicevox/song.json} +0 -0
- /package/examples/{speech-kana.json → voicevox/speech-kana.json} +0 -0
- /package/examples/{speech.json → voicevox/speech.json} +0 -0
|
@@ -0,0 +1,354 @@
|
|
|
1
|
+
import { invalid } from "../errors.js";
|
|
2
|
+
import { MAX_F0_HZ, MIN_F0_HZ } from "../limits.js";
|
|
3
|
+
import { isUnvoicedVowel } from "../text/mora.js";
|
|
4
|
+
import { boundaryIsSentenceEnd, formatPhraseBody } from "../text/notation.js";
|
|
5
|
+
import { ENGINE_VERSION, PLAN_VERSION } from "../version.js";
|
|
6
|
+
import { compileSegments, phoneDraft } from "./acoustics.js";
|
|
7
|
+
import { PHONEMES } from "./phonemes.js";
|
|
8
|
+
import { FUJISAKI_ALPHA, FUJISAKI_BETA, FUJISAKI_GAMMA, applyPitchScale, evaluateFujisakiHz } from "./prosody.js";
|
|
9
|
+
export const SPEECH_TIMING = {
|
|
10
|
+
mora: 0.108,
|
|
11
|
+
consonantShare: 0.6,
|
|
12
|
+
minVowel: 0.04,
|
|
13
|
+
moraicNasal: 0.08,
|
|
14
|
+
closure: 0.08,
|
|
15
|
+
longVowel: 0.1,
|
|
16
|
+
devoiced: 0.055,
|
|
17
|
+
pausePhrase: 0,
|
|
18
|
+
pauseComma: 0.2,
|
|
19
|
+
pauseSentence: 0.38,
|
|
20
|
+
pauseParagraph: 0.7,
|
|
21
|
+
phraseFinalStretch: 1.12,
|
|
22
|
+
sentenceFinalStretch: 1.3,
|
|
23
|
+
questionStretch: 1.45,
|
|
24
|
+
sentenceInitialStretch: 1.05,
|
|
25
|
+
};
|
|
26
|
+
const ACCENT_LEAD_SECONDS = 0.08;
|
|
27
|
+
export function boundaryPauseSeconds(boundary, timing, paragraphAfter = false) {
|
|
28
|
+
const speedFactor = 1 / timing.speed ** 0.7;
|
|
29
|
+
let base;
|
|
30
|
+
switch (boundary) {
|
|
31
|
+
case "pause":
|
|
32
|
+
base = timing.pauseLength ?? SPEECH_TIMING.pauseComma;
|
|
33
|
+
break;
|
|
34
|
+
case "sentence":
|
|
35
|
+
case "question":
|
|
36
|
+
case "exclamation":
|
|
37
|
+
base =
|
|
38
|
+
timing.pauseLength === undefined
|
|
39
|
+
? SPEECH_TIMING.pauseSentence
|
|
40
|
+
: Math.max(timing.pauseLength, SPEECH_TIMING.pauseSentence);
|
|
41
|
+
break;
|
|
42
|
+
case "paragraph":
|
|
43
|
+
base = SPEECH_TIMING.pauseParagraph;
|
|
44
|
+
break;
|
|
45
|
+
default:
|
|
46
|
+
base = 0;
|
|
47
|
+
}
|
|
48
|
+
if (paragraphAfter && boundary !== "phrase" && boundary !== "end")
|
|
49
|
+
base = Math.max(base, SPEECH_TIMING.pauseParagraph);
|
|
50
|
+
return base * timing.pauseScale * speedFactor;
|
|
51
|
+
}
|
|
52
|
+
export function utteranceFinalPauseSeconds(boundary, timing) {
|
|
53
|
+
return boundary === "pause" ? boundaryPauseSeconds(boundary, timing) : 0;
|
|
54
|
+
}
|
|
55
|
+
function sentenceExclamationFlags(reading) {
|
|
56
|
+
const flags = new Array(reading.phrases.length).fill(false);
|
|
57
|
+
let groupStart = 0;
|
|
58
|
+
for (const [index, phrase] of reading.phrases.entries()) {
|
|
59
|
+
if (!boundaryIsSentenceEnd(phrase.boundary) && index !== reading.phrases.length - 1)
|
|
60
|
+
continue;
|
|
61
|
+
let exclamatory = false;
|
|
62
|
+
for (let k = groupStart; k <= index; k++)
|
|
63
|
+
if (reading.phrases[k]?.exclamatory === true)
|
|
64
|
+
exclamatory = true;
|
|
65
|
+
for (let k = groupStart; k <= index; k++)
|
|
66
|
+
flags[k] = exclamatory;
|
|
67
|
+
groupStart = index + 1;
|
|
68
|
+
}
|
|
69
|
+
return flags;
|
|
70
|
+
}
|
|
71
|
+
export function planSpeech(input) {
|
|
72
|
+
const path = input.path ?? "$.text";
|
|
73
|
+
const { reading, voice, sampleRate } = input;
|
|
74
|
+
const rate = input.speed * voice.rateScale;
|
|
75
|
+
const phones = [];
|
|
76
|
+
const warnings = [...reading.warnings];
|
|
77
|
+
const moraTimes = [];
|
|
78
|
+
const phraseTimes = [];
|
|
79
|
+
const moraOnsets = [];
|
|
80
|
+
const moraEnds = [];
|
|
81
|
+
const timing = { speed: input.speed, pauseLength: input.pauseLength, pauseScale: input.pauseScale };
|
|
82
|
+
const exclamatoryFlags = sentenceExclamationFlags(reading);
|
|
83
|
+
const lastPhraseIndex = reading.phrases.length - 1;
|
|
84
|
+
const paragraphAfter = (phraseIndex) => reading.sentences.some((sentence) => sentence.paragraphEnd && sentence.phraseEnd - 1 === phraseIndex);
|
|
85
|
+
let clock = 0;
|
|
86
|
+
const pushPhone = (phone) => {
|
|
87
|
+
phones.push(phone);
|
|
88
|
+
clock += phone.seconds;
|
|
89
|
+
};
|
|
90
|
+
const rest = (seconds) => {
|
|
91
|
+
if (seconds <= 0)
|
|
92
|
+
return;
|
|
93
|
+
pushPhone({ ...phoneDraft("pau", seconds), isVowelOfMora: false, moraStart: false });
|
|
94
|
+
};
|
|
95
|
+
rest(input.prePause);
|
|
96
|
+
let moraCounter = 0;
|
|
97
|
+
let sentenceStart = true;
|
|
98
|
+
for (const [phraseIndex, phrase] of reading.phrases.entries()) {
|
|
99
|
+
const phraseStart = clock;
|
|
100
|
+
const onsets = [];
|
|
101
|
+
const ends = [];
|
|
102
|
+
const count = phrase.moras.length;
|
|
103
|
+
const sentenceEnd = boundaryIsSentenceEnd(phrase.boundary);
|
|
104
|
+
const phraseEnd = phrase.boundary !== "phrase";
|
|
105
|
+
for (const [moraIndex, mora] of phrase.moras.entries()) {
|
|
106
|
+
const moraStartClock = clock;
|
|
107
|
+
const last = moraIndex === count - 1;
|
|
108
|
+
const first = moraIndex === 0;
|
|
109
|
+
const isLong = mora.text === "ー";
|
|
110
|
+
const devoiced = isUnvoicedVowel(mora.vowel);
|
|
111
|
+
let stretch = 1;
|
|
112
|
+
if (last && sentenceEnd)
|
|
113
|
+
stretch *= phrase.boundary === "question" ? SPEECH_TIMING.questionStretch : SPEECH_TIMING.sentenceFinalStretch;
|
|
114
|
+
else if (last && phraseEnd)
|
|
115
|
+
stretch *= SPEECH_TIMING.phraseFinalStretch;
|
|
116
|
+
if (first && sentenceStart)
|
|
117
|
+
stretch *= SPEECH_TIMING.sentenceInitialStretch;
|
|
118
|
+
if (phrase.accent === moraIndex + 1)
|
|
119
|
+
stretch *= 1.04;
|
|
120
|
+
const emphasis = exclamatoryFlags[phraseIndex] === true ? 1 : 0;
|
|
121
|
+
const phonemes = [];
|
|
122
|
+
let consonantSeconds = 0;
|
|
123
|
+
if (mora.consonant !== null) {
|
|
124
|
+
const spec = PHONEMES[mora.consonant];
|
|
125
|
+
consonantSeconds = spec.duration / rate;
|
|
126
|
+
phonemes.push(mora.consonant);
|
|
127
|
+
pushPhone({
|
|
128
|
+
...phoneDraft(mora.consonant, consonantSeconds, {
|
|
129
|
+
moraIndex: moraCounter,
|
|
130
|
+
phraseIndex,
|
|
131
|
+
boundaryBefore: first && (phraseIndex === 0 || sentenceStart || phones.at(-1)?.phoneme === "pau"),
|
|
132
|
+
emphasis,
|
|
133
|
+
}),
|
|
134
|
+
isVowelOfMora: false,
|
|
135
|
+
moraStart: true,
|
|
136
|
+
});
|
|
137
|
+
}
|
|
138
|
+
let vowelSeconds;
|
|
139
|
+
if (mora.vowel === "N")
|
|
140
|
+
vowelSeconds = SPEECH_TIMING.moraicNasal;
|
|
141
|
+
else if (mora.vowel === "cl")
|
|
142
|
+
vowelSeconds = SPEECH_TIMING.closure;
|
|
143
|
+
else if (isLong)
|
|
144
|
+
vowelSeconds = SPEECH_TIMING.longVowel;
|
|
145
|
+
else if (devoiced)
|
|
146
|
+
vowelSeconds = SPEECH_TIMING.devoiced;
|
|
147
|
+
else
|
|
148
|
+
vowelSeconds = Math.max(SPEECH_TIMING.minVowel, SPEECH_TIMING.mora - SPEECH_TIMING.consonantShare * consonantSeconds * rate);
|
|
149
|
+
vowelSeconds = (vowelSeconds * stretch) / rate;
|
|
150
|
+
const previous = phones.at(-1);
|
|
151
|
+
const continuing = isLong && previous !== undefined && previous.phoneme === mora.vowel;
|
|
152
|
+
phonemes.push(mora.vowel);
|
|
153
|
+
pushPhone({
|
|
154
|
+
...phoneDraft(mora.vowel, vowelSeconds, {
|
|
155
|
+
moraIndex: moraCounter,
|
|
156
|
+
phraseIndex,
|
|
157
|
+
articulation: continuing ? "continue" : "onset",
|
|
158
|
+
boundaryBefore: mora.consonant === null && first && (phraseIndex === 0 || sentenceStart || previous?.phoneme === "pau"),
|
|
159
|
+
boundaryAfter: last && phraseEnd,
|
|
160
|
+
emphasis,
|
|
161
|
+
gainDb: last && sentenceEnd ? -2 : 0,
|
|
162
|
+
rdOffset: input.breathiness,
|
|
163
|
+
}),
|
|
164
|
+
isVowelOfMora: true,
|
|
165
|
+
moraStart: mora.consonant === null,
|
|
166
|
+
});
|
|
167
|
+
onsets.push(clock - vowelSeconds);
|
|
168
|
+
ends.push(clock);
|
|
169
|
+
moraTimes.push({ start: moraStartClock, end: clock, phraseIndex, text: mora.text, phonemes });
|
|
170
|
+
moraCounter += 1;
|
|
171
|
+
}
|
|
172
|
+
moraOnsets.push(onsets);
|
|
173
|
+
moraEnds.push(ends);
|
|
174
|
+
phraseTimes.push({ start: phraseStart, end: clock });
|
|
175
|
+
rest(phraseIndex === lastPhraseIndex
|
|
176
|
+
? utteranceFinalPauseSeconds(phrase.boundary, timing)
|
|
177
|
+
: boundaryPauseSeconds(phrase.boundary, timing, paragraphAfter(phraseIndex)));
|
|
178
|
+
sentenceStart = sentenceEnd;
|
|
179
|
+
}
|
|
180
|
+
rest(input.postPause);
|
|
181
|
+
const baseHz = voice.baseF0 * 0.82 * 2 ** (input.pitchSemitones / 12);
|
|
182
|
+
const intonation = input.intonation * voice.intonationScale;
|
|
183
|
+
const phraseCommands = [];
|
|
184
|
+
const accentCommands = [];
|
|
185
|
+
const tones = [];
|
|
186
|
+
let newSentence = true;
|
|
187
|
+
for (const [phraseIndex, phrase] of reading.phrases.entries()) {
|
|
188
|
+
const times = phraseTimes[phraseIndex];
|
|
189
|
+
const onsets = moraOnsets[phraseIndex] ?? [];
|
|
190
|
+
const ends = moraEnds[phraseIndex] ?? [];
|
|
191
|
+
if (times === undefined)
|
|
192
|
+
continue;
|
|
193
|
+
const exclaim = exclamatoryFlags[phraseIndex] === true ? 1.2 : 1;
|
|
194
|
+
if (newSentence) {
|
|
195
|
+
phraseCommands.push({ at: times.start - 0.12, amplitude: 0.28 * intonation * exclaim });
|
|
196
|
+
}
|
|
197
|
+
else {
|
|
198
|
+
const previous = reading.phrases[phraseIndex - 1];
|
|
199
|
+
if (previous !== undefined && previous.boundary === "pause") {
|
|
200
|
+
phraseCommands.push({ at: times.start - 0.1, amplitude: 0.16 * intonation * exclaim });
|
|
201
|
+
}
|
|
202
|
+
else {
|
|
203
|
+
phraseCommands.push({ at: times.start - 0.08, amplitude: 0.05 * intonation });
|
|
204
|
+
}
|
|
205
|
+
}
|
|
206
|
+
const count = phrase.moras.length;
|
|
207
|
+
const accent = Math.min(phrase.accent, count);
|
|
208
|
+
const firstOnset = onsets[0] ?? times.start;
|
|
209
|
+
const secondOnset = onsets[1] ?? firstOnset;
|
|
210
|
+
if (count >= 1) {
|
|
211
|
+
if (accent === 1) {
|
|
212
|
+
accentCommands.push({
|
|
213
|
+
start: firstOnset - ACCENT_LEAD_SECONDS,
|
|
214
|
+
end: (ends[0] ?? times.end) - 0.03,
|
|
215
|
+
amplitude: 0.44 * intonation * exclaim,
|
|
216
|
+
});
|
|
217
|
+
}
|
|
218
|
+
else if (accent > 1) {
|
|
219
|
+
accentCommands.push({
|
|
220
|
+
start: secondOnset - ACCENT_LEAD_SECONDS,
|
|
221
|
+
end: (ends[accent - 1] ?? times.end) - 0.03,
|
|
222
|
+
amplitude: 0.4 * intonation * exclaim,
|
|
223
|
+
});
|
|
224
|
+
}
|
|
225
|
+
else if (count >= 2) {
|
|
226
|
+
accentCommands.push({
|
|
227
|
+
start: secondOnset - ACCENT_LEAD_SECONDS,
|
|
228
|
+
end: times.end - 0.02,
|
|
229
|
+
amplitude: 0.32 * intonation * exclaim,
|
|
230
|
+
});
|
|
231
|
+
}
|
|
232
|
+
}
|
|
233
|
+
const sentenceEnd = boundaryIsSentenceEnd(phrase.boundary);
|
|
234
|
+
if (sentenceEnd) {
|
|
235
|
+
const lastOnset = onsets[count - 1] ?? times.start;
|
|
236
|
+
if (phrase.boundary === "question" && input.upspeak) {
|
|
237
|
+
tones.push({ start: lastOnset, end: times.end, amount: 0.38 * Math.max(0.4, intonation) });
|
|
238
|
+
}
|
|
239
|
+
else if (phrase.boundary !== "question") {
|
|
240
|
+
tones.push({
|
|
241
|
+
start: Math.max(times.start, times.end - 0.25),
|
|
242
|
+
end: times.end,
|
|
243
|
+
amount: -0.12 * Math.max(0.3, intonation),
|
|
244
|
+
});
|
|
245
|
+
}
|
|
246
|
+
}
|
|
247
|
+
newSentence = sentenceEnd;
|
|
248
|
+
}
|
|
249
|
+
const model = {
|
|
250
|
+
baseHz,
|
|
251
|
+
alpha: FUJISAKI_ALPHA,
|
|
252
|
+
beta: FUJISAKI_BETA,
|
|
253
|
+
gamma: FUJISAKI_GAMMA,
|
|
254
|
+
phrases: phraseCommands,
|
|
255
|
+
accents: accentCommands,
|
|
256
|
+
tones,
|
|
257
|
+
};
|
|
258
|
+
const totalSeconds = clock;
|
|
259
|
+
const step = 0.005;
|
|
260
|
+
const pitch = [];
|
|
261
|
+
let f0Min = Infinity;
|
|
262
|
+
let f0Max = 0;
|
|
263
|
+
for (let t = 0; t <= totalSeconds + step; t += step) {
|
|
264
|
+
const hz = applyPitchScale(evaluateFujisakiHz(model, t), input.pitchScale);
|
|
265
|
+
pitch.push({ sample: Math.round(t * sampleRate), hz });
|
|
266
|
+
}
|
|
267
|
+
for (const [index, phone] of phones.entries()) {
|
|
268
|
+
if (!PHONEMES[phone.phoneme].voiced)
|
|
269
|
+
continue;
|
|
270
|
+
let start = 0;
|
|
271
|
+
for (let k = 0; k < index; k++)
|
|
272
|
+
start += phones[k]?.seconds ?? 0;
|
|
273
|
+
const end = start + phone.seconds;
|
|
274
|
+
for (const point of pitch) {
|
|
275
|
+
const t = point.sample / sampleRate;
|
|
276
|
+
if (t < start || t > end)
|
|
277
|
+
continue;
|
|
278
|
+
f0Min = Math.min(f0Min, point.hz);
|
|
279
|
+
f0Max = Math.max(f0Max, point.hz);
|
|
280
|
+
}
|
|
281
|
+
}
|
|
282
|
+
if (!Number.isFinite(f0Min)) {
|
|
283
|
+
f0Min = baseHz;
|
|
284
|
+
f0Max = baseHz;
|
|
285
|
+
}
|
|
286
|
+
if (f0Max > MAX_F0_HZ || f0Max > sampleRate * 0.2 || f0Min < MIN_F0_HZ) {
|
|
287
|
+
invalid(input.pitchScale === undefined ? "$.pitchSemitones" : "$.pitch", `The speech contour reaches ${f0Max.toFixed(0)} Hz (minimum ${f0Min.toFixed(0)} Hz); the engine synthesizes ${MIN_F0_HZ}–${Math.min(MAX_F0_HZ, sampleRate * 0.2).toFixed(0)} Hz at ${sampleRate} Hz.`, {
|
|
288
|
+
code: "PITCH_OUT_OF_RANGE",
|
|
289
|
+
hint: "Lower pitchSemitones (or pitch), reduce intonation, or raise sampleRate.",
|
|
290
|
+
repairOptions: [
|
|
291
|
+
{ action: "lower-pitch", description: "Reduce pitchSemitones or pitch.", path: "$.pitchSemitones" },
|
|
292
|
+
{ action: "raise-sample-rate", description: "Use a higher sampleRate such as 24000.", path: "$.sampleRate" },
|
|
293
|
+
],
|
|
294
|
+
});
|
|
295
|
+
}
|
|
296
|
+
if (f0Max > voice.f0Max || f0Min < voice.f0Min) {
|
|
297
|
+
warnings.push({
|
|
298
|
+
severity: "warning",
|
|
299
|
+
code: "F0_OUTSIDE_VOICE_RANGE",
|
|
300
|
+
message: `The contour spans ${f0Min.toFixed(0)}–${f0Max.toFixed(0)} Hz; voice ${voice.id} is tuned for ${voice.f0Min}–${voice.f0Max} Hz.`,
|
|
301
|
+
help: "Choose another voice or reduce the pitch shift.",
|
|
302
|
+
path,
|
|
303
|
+
});
|
|
304
|
+
}
|
|
305
|
+
const compiled = compileSegments(phones, voice, sampleRate);
|
|
306
|
+
const boundaries = compiled.boundaries;
|
|
307
|
+
const sampleOf = (seconds) => Math.round(seconds * sampleRate);
|
|
308
|
+
const moras = moraTimes.map((mora, index) => ({
|
|
309
|
+
index,
|
|
310
|
+
text: mora.text,
|
|
311
|
+
start: Math.min(compiled.frames, sampleOf(mora.start)),
|
|
312
|
+
end: Math.min(compiled.frames, sampleOf(mora.end)),
|
|
313
|
+
phraseIndex: mora.phraseIndex,
|
|
314
|
+
phonemes: mora.phonemes,
|
|
315
|
+
}));
|
|
316
|
+
const phrases = phraseTimes.map((times, index) => {
|
|
317
|
+
const phrase = reading.phrases[index];
|
|
318
|
+
return {
|
|
319
|
+
index,
|
|
320
|
+
text: phrase === undefined ? "" : phrase.moras.map((m) => m.text).join(""),
|
|
321
|
+
kana: phrase === undefined ? "" : formatPhraseBody(phrase.moras, phrase.accent),
|
|
322
|
+
accent: phrase?.accent ?? 0,
|
|
323
|
+
start: Math.min(compiled.frames, sampleOf(times.start)),
|
|
324
|
+
end: Math.min(compiled.frames, sampleOf(times.end)),
|
|
325
|
+
boundary: phrase?.boundary ?? "end",
|
|
326
|
+
};
|
|
327
|
+
});
|
|
328
|
+
void boundaries;
|
|
329
|
+
return {
|
|
330
|
+
planVersion: PLAN_VERSION,
|
|
331
|
+
engineVersion: ENGINE_VERSION,
|
|
332
|
+
kind: "speech",
|
|
333
|
+
sampleRate,
|
|
334
|
+
seed: input.seed,
|
|
335
|
+
frames: compiled.frames,
|
|
336
|
+
voice,
|
|
337
|
+
volume: input.volume,
|
|
338
|
+
jitter: input.jitter ?? voice.jitter,
|
|
339
|
+
shimmerDb: voice.shimmerDb,
|
|
340
|
+
flutter: voice.flutter,
|
|
341
|
+
segments: compiled.segments,
|
|
342
|
+
pitch,
|
|
343
|
+
vibrato: [],
|
|
344
|
+
moras,
|
|
345
|
+
phrases,
|
|
346
|
+
notes: [],
|
|
347
|
+
reading,
|
|
348
|
+
warnings,
|
|
349
|
+
adjustments: [],
|
|
350
|
+
f0Min,
|
|
351
|
+
f0Max,
|
|
352
|
+
};
|
|
353
|
+
}
|
|
354
|
+
//# sourceMappingURL=speech.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"speech.js","sourceRoot":"","sources":["../../src/synth/speech.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,OAAO,EAAE,MAAM,cAAc,CAAC;AACvC,OAAO,EAAE,SAAS,EAAE,SAAS,EAAE,MAAM,cAAc,CAAC;AAEpD,OAAO,EAAE,eAAe,EAAE,MAAM,iBAAiB,CAAC;AAElD,OAAO,EAAE,qBAAqB,EAAE,gBAAgB,EAAE,MAAM,qBAAqB,CAAC;AAE9E,OAAO,EAAE,cAAc,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAE7D,OAAO,EAAE,eAAe,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAC7D,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAGzC,OAAO,EAAE,cAAc,EAAE,aAAa,EAAE,cAAc,EAAE,eAAe,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAC;AAuBlH,MAAM,CAAC,MAAM,aAAa,GAgBtB;IACF,IAAI,EAAE,KAAK;IACX,cAAc,EAAE,GAAG;IACnB,QAAQ,EAAE,IAAI;IACd,WAAW,EAAE,IAAI;IACjB,OAAO,EAAE,IAAI;IACb,SAAS,EAAE,GAAG;IACd,QAAQ,EAAE,KAAK;IACf,WAAW,EAAE,CAAC;IACd,UAAU,EAAE,GAAG;IACf,aAAa,EAAE,IAAI;IACnB,cAAc,EAAE,GAAG;IACnB,kBAAkB,EAAE,IAAI;IACxB,oBAAoB,EAAE,GAAG;IACzB,eAAe,EAAE,IAAI;IACrB,sBAAsB,EAAE,IAAI;CAC7B,CAAC;AAaF,MAAM,mBAAmB,GAAG,IAAI,CAAC;AAEjC,MAAM,UAAU,oBAAoB,CAClC,QAAsB,EACtB,MAAmB,EACnB,iBAA0B,KAAK;IAE/B,MAAM,WAAW,GAAG,CAAC,GAAG,MAAM,CAAC,KAAK,IAAI,GAAG,CAAC;IAC5C,IAAI,IAAY,CAAC;IACjB,QAAQ,QAAQ,EAAE,CAAC;QACjB,KAAK,OAAO;YACV,IAAI,GAAG,MAAM,CAAC,WAAW,IAAI,aAAa,CAAC,UAAU,CAAC;YACtD,MAAM;QACR,KAAK,UAAU,CAAC;QAChB,KAAK,UAAU,CAAC;QAChB,KAAK,aAAa;YAChB,IAAI;gBACF,MAAM,CAAC,WAAW,KAAK,SAAS;oBAC9B,CAAC,CAAC,aAAa,CAAC,aAAa;oBAC7B,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,WAAW,EAAE,aAAa,CAAC,aAAa,CAAC,CAAC;YAChE,MAAM;QACR,KAAK,WAAW;YACd,IAAI,GAAG,aAAa,CAAC,cAAc,CAAC;YACpC,MAAM;QACR;YACE,IAAI,GAAG,CAAC,CAAC;IACb,CAAC;IACD,IAAI,cAAc,IAAI,QAAQ,KAAK,QAAQ,IAAI,QAAQ,KAAK,KAAK;QAC/D,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,aAAa,CAAC,cAAc,CAAC,CAAC;IACtD,OAAO,IAAI,GAAG,MAAM,CAAC,UAAU,GAAG,WAAW,CAAC;AAChD,CAAC;AAED,MAAM,UAAU,0BAA0B,CAAC,QAAsB,EAAE,MAAmB;IACpF,OAAO,QAAQ,KAAK,OAAO,CAAC,CAAC,CAAC,oBAAoB,CAAC,QAAQ,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AAC3E,CAAC;AAED,SAAS,wBAAwB,CAAC,OAAoB;IACpD,MAAM,KAAK,GAAc,IAAI,KAAK,CAAU,OAAO,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAChF,IAAI,UAAU,GAAG,CAAC,CAAC;IACnB,KAAK,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;QACxD,IAAI,CAAC,qBAAqB,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,KAAK,KAAK,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC;YAAE,SAAS;QAC9F,IAAI,WAAW,GAAG,KAAK,CAAC;QACxB,KAAK,IAAI,CAAC,GAAG,UAAU,EAAE,CAAC,IAAI,KAAK,EAAE,CAAC,EAAE;YAAE,IAAI,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,WAAW,KAAK,IAAI;gBAAE,WAAW,GAAG,IAAI,CAAC;QAC3G,KAAK,IAAI,CAAC,GAAG,UAAU,EAAE,CAAC,IAAI,KAAK,EAAE,CAAC,EAAE;YAAE,KAAK,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC;QACjE,UAAU,GAAG,KAAK,GAAG,CAAC,CAAC;IACzB,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,KAAsB;IAC/C,MAAM,IAAI,GAAG,KAAK,CAAC,IAAI,IAAI,QAAQ,CAAC;IACpC,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,UAAU,EAAE,GAAG,KAAK,CAAC;IAC7C,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,SAAS,CAAC;IAC3C,MAAM,MAAM,GAAY,EAAE,CAAC;IAC3B,MAAM,QAAQ,GAAiB,CAAC,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAC;IACrD,MAAM,SAAS,GAA6F,EAAE,CAAC;IAC/G,MAAM,WAAW,GAAqC,EAAE,CAAC;IACzD,MAAM,UAAU,GAAe,EAAE,CAAC;IAClC,MAAM,QAAQ,GAAe,EAAE,CAAC;IAChC,MAAM,MAAM,GAAgB,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,WAAW,EAAE,KAAK,CAAC,WAAW,EAAE,UAAU,EAAE,KAAK,CAAC,UAAU,EAAE,CAAC;IACjH,MAAM,gBAAgB,GAAG,wBAAwB,CAAC,OAAO,CAAC,CAAC;IAC3D,MAAM,eAAe,GAAG,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC;IACnD,MAAM,cAAc,GAAG,CAAC,WAAmB,EAAW,EAAE,CACtD,OAAO,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,YAAY,IAAI,QAAQ,CAAC,SAAS,GAAG,CAAC,KAAK,WAAW,CAAC,CAAC;IACxG,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,MAAM,SAAS,GAAG,CAAC,KAAY,EAAQ,EAAE;QACvC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACnB,KAAK,IAAI,KAAK,CAAC,OAAO,CAAC;IACzB,CAAC,CAAC;IACF,MAAM,IAAI,GAAG,CAAC,OAAe,EAAQ,EAAE;QACrC,IAAI,OAAO,IAAI,CAAC;YAAE,OAAO;QACzB,SAAS,CAAC,EAAE,GAAG,UAAU,CAAC,KAAK,EAAE,OAAO,CAAC,EAAE,aAAa,EAAE,KAAK,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;IACvF,CAAC,CAAC;IACF,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IACrB,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,IAAI,aAAa,GAAG,IAAI,CAAC;IACzB,KAAK,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;QAC9D,MAAM,WAAW,GAAG,KAAK,CAAC;QAC1B,MAAM,MAAM,GAAa,EAAE,CAAC;QAC5B,MAAM,IAAI,GAAa,EAAE,CAAC;QAC1B,MAAM,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC;QAClC,MAAM,WAAW,GAAG,qBAAqB,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;QAC3D,MAAM,SAAS,GAAG,MAAM,CAAC,QAAQ,KAAK,QAAQ,CAAC;QAC/C,KAAK,MAAM,CAAC,SAAS,EAAE,IAAI,CAAC,IAAI,MAAM,CAAC,KAAK,CAAC,OAAO,EAAE,EAAE,CAAC;YACvD,MAAM,cAAc,GAAG,KAAK,CAAC;YAC7B,MAAM,IAAI,GAAG,SAAS,KAAK,KAAK,GAAG,CAAC,CAAC;YACrC,MAAM,KAAK,GAAG,SAAS,KAAK,CAAC,CAAC;YAC9B,MAAM,MAAM,GAAG,IAAI,CAAC,IAAI,KAAK,GAAG,CAAC;YACjC,MAAM,QAAQ,GAAG,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAC7C,IAAI,OAAO,GAAG,CAAC,CAAC;YAChB,IAAI,IAAI,IAAI,WAAW;gBACrB,OAAO,IAAI,MAAM,CAAC,QAAQ,KAAK,UAAU,CAAC,CAAC,CAAC,aAAa,CAAC,eAAe,CAAC,CAAC,CAAC,aAAa,CAAC,oBAAoB,CAAC;iBAC5G,IAAI,IAAI,IAAI,SAAS;gBAAE,OAAO,IAAI,aAAa,CAAC,kBAAkB,CAAC;YACxE,IAAI,KAAK,IAAI,aAAa;gBAAE,OAAO,IAAI,aAAa,CAAC,sBAAsB,CAAC;YAC5E,IAAI,MAAM,CAAC,MAAM,KAAK,SAAS,GAAG,CAAC;gBAAE,OAAO,IAAI,IAAI,CAAC;YACrD,MAAM,QAAQ,GAAG,gBAAgB,CAAC,WAAW,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;YAChE,MAAM,QAAQ,GAAc,EAAE,CAAC;YAC/B,IAAI,gBAAgB,GAAG,CAAC,CAAC;YACzB,IAAI,IAAI,CAAC,SAAS,KAAK,IAAI,EAAE,CAAC;gBAC5B,MAAM,IAAI,GAAG,QAAQ,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;gBACtC,gBAAgB,GAAG,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC;gBACxC,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;gBAC9B,SAAS,CAAC;oBACR,GAAG,UAAU,CAAC,IAAI,CAAC,SAAS,EAAE,gBAAgB,EAAE;wBAC9C,SAAS,EAAE,WAAW;wBACtB,WAAW;wBACX,cAAc,EAAE,KAAK,IAAI,CAAC,WAAW,KAAK,CAAC,IAAI,aAAa,IAAI,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,OAAO,KAAK,KAAK,CAAC;wBACjG,QAAQ;qBACT,CAAC;oBACF,aAAa,EAAE,KAAK;oBACpB,SAAS,EAAE,IAAI;iBAChB,CAAC,CAAC;YACL,CAAC;YACD,IAAI,YAAoB,CAAC;YACzB,IAAI,IAAI,CAAC,KAAK,KAAK,GAAG;gBAAE,YAAY,GAAG,aAAa,CAAC,WAAW,CAAC;iBAC5D,IAAI,IAAI,CAAC,KAAK,KAAK,IAAI;gBAAE,YAAY,GAAG,aAAa,CAAC,OAAO,CAAC;iBAC9D,IAAI,MAAM;gBAAE,YAAY,GAAG,aAAa,CAAC,SAAS,CAAC;iBACnD,IAAI,QAAQ;gBAAE,YAAY,GAAG,aAAa,CAAC,QAAQ,CAAC;;gBAEvD,YAAY,GAAG,IAAI,CAAC,GAAG,CACrB,aAAa,CAAC,QAAQ,EACtB,aAAa,CAAC,IAAI,GAAG,aAAa,CAAC,cAAc,GAAG,gBAAgB,GAAG,IAAI,CAC5E,CAAC;YACJ,YAAY,GAAG,CAAC,YAAY,GAAG,OAAO,CAAC,GAAG,IAAI,CAAC;YAC/C,MAAM,QAAQ,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;YAC/B,MAAM,UAAU,GAAG,MAAM,IAAI,QAAQ,KAAK,SAAS,IAAI,QAAQ,CAAC,OAAO,KAAK,IAAI,CAAC,KAAK,CAAC;YACvF,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YAC1B,SAAS,CAAC;gBACR,GAAG,UAAU,CAAC,IAAI,CAAC,KAAK,EAAE,YAAY,EAAE;oBACtC,SAAS,EAAE,WAAW;oBACtB,WAAW;oBACX,YAAY,EAAE,UAAU,CAAC,CAAC,CAAC,UAAU,CAAC,CAAC,CAAC,OAAO;oBAC/C,cAAc,EACZ,IAAI,CAAC,SAAS,KAAK,IAAI,IAAI,KAAK,IAAI,CAAC,WAAW,KAAK,CAAC,IAAI,aAAa,IAAI,QAAQ,EAAE,OAAO,KAAK,KAAK,CAAC;oBACzG,aAAa,EAAE,IAAI,IAAI,SAAS;oBAChC,QAAQ;oBACR,MAAM,EAAE,IAAI,IAAI,WAAW,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;oBACpC,QAAQ,EAAE,KAAK,CAAC,WAAW;iBAC5B,CAAC;gBACF,aAAa,EAAE,IAAI;gBACnB,SAAS,EAAE,IAAI,CAAC,SAAS,KAAK,IAAI;aACnC,CAAC,CAAC;YACH,MAAM,CAAC,IAAI,CAAC,KAAK,GAAG,YAAY,CAAC,CAAC;YAClC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACjB,SAAS,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,cAAc,EAAE,GAAG,EAAE,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE,IAAI,CAAC,IAAI,EAAE,QAAQ,EAAE,CAAC,CAAC;YAC9F,WAAW,IAAI,CAAC,CAAC;QACnB,CAAC;QACD,UAAU,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QACxB,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACpB,WAAW,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,WAAW,EAAE,GAAG,EAAE,KAAK,EAAE,CAAC,CAAC;QACrD,IAAI,CACF,WAAW,KAAK,eAAe;YAC7B,CAAC,CAAC,0BAA0B,CAAC,MAAM,CAAC,QAAQ,EAAE,MAAM,CAAC;YACrD,CAAC,CAAC,oBAAoB,CAAC,MAAM,CAAC,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,WAAW,CAAC,CAAC,CAC/E,CAAC;QACF,aAAa,GAAG,WAAW,CAAC;IAC9B,CAAC;IACD,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC;IACtB,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,GAAG,IAAI,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,cAAc,GAAG,EAAE,CAAC,CAAC;IACtE,MAAM,UAAU,GAAG,KAAK,CAAC,UAAU,GAAG,KAAK,CAAC,eAAe,CAAC;IAC5D,MAAM,cAAc,GAAoB,EAAE,CAAC;IAC3C,MAAM,cAAc,GAAoB,EAAE,CAAC;IAC3C,MAAM,KAAK,GAAmB,EAAE,CAAC;IACjC,IAAI,WAAW,GAAG,IAAI,CAAC;IACvB,KAAK,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,IAAI,OAAO,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;QAC9D,MAAM,KAAK,GAAG,WAAW,CAAC,WAAW,CAAC,CAAC;QACvC,MAAM,MAAM,GAAG,UAAU,CAAC,WAAW,CAAC,IAAI,EAAE,CAAC;QAC7C,MAAM,IAAI,GAAG,QAAQ,CAAC,WAAW,CAAC,IAAI,EAAE,CAAC;QACzC,IAAI,KAAK,KAAK,SAAS;YAAE,SAAS;QAClC,MAAM,OAAO,GAAG,gBAAgB,CAAC,WAAW,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC;QACjE,IAAI,WAAW,EAAE,CAAC;YAChB,cAAc,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,KAAK,CAAC,KAAK,GAAG,IAAI,EAAE,SAAS,EAAE,IAAI,GAAG,UAAU,GAAG,OAAO,EAAE,CAAC,CAAC;QAC1F,CAAC;aAAM,CAAC;YACN,MAAM,QAAQ,GAAG,OAAO,CAAC,OAAO,CAAC,WAAW,GAAG,CAAC,CAAC,CAAC;YAClD,IAAI,QAAQ,KAAK,SAAS,IAAI,QAAQ,CAAC,QAAQ,KAAK,OAAO,EAAE,CAAC;gBAC5D,cAAc,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,KAAK,CAAC,KAAK,GAAG,GAAG,EAAE,SAAS,EAAE,IAAI,GAAG,UAAU,GAAG,OAAO,EAAE,CAAC,CAAC;YACzF,CAAC;iBAAM,CAAC;gBACN,cAAc,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,KAAK,CAAC,KAAK,GAAG,IAAI,EAAE,SAAS,EAAE,IAAI,GAAG,UAAU,EAAE,CAAC,CAAC;YAChF,CAAC;QACH,CAAC;QACD,MAAM,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,MAAM,CAAC;QAClC,MAAM,MAAM,GAAG,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;QAC9C,MAAM,UAAU,GAAG,MAAM,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC;QAC5C,MAAM,WAAW,GAAG,MAAM,CAAC,CAAC,CAAC,IAAI,UAAU,CAAC;QAC5C,IAAI,KAAK,IAAI,CAAC,EAAE,CAAC;YACf,IAAI,MAAM,KAAK,CAAC,EAAE,CAAC;gBACjB,cAAc,CAAC,IAAI,CAAC;oBAClB,KAAK,EAAE,UAAU,GAAG,mBAAmB;oBACvC,GAAG,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,GAAG,CAAC,GAAG,IAAI;oBAClC,SAAS,EAAE,IAAI,GAAG,UAAU,GAAG,OAAO;iBACvC,CAAC,CAAC;YACL,CAAC;iBAAM,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;gBACtB,cAAc,CAAC,IAAI,CAAC;oBAClB,KAAK,EAAE,WAAW,GAAG,mBAAmB;oBACxC,GAAG,EAAE,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,KAAK,CAAC,GAAG,CAAC,GAAG,IAAI;oBAC3C,SAAS,EAAE,GAAG,GAAG,UAAU,GAAG,OAAO;iBACtC,CAAC,CAAC;YACL,CAAC;iBAAM,IAAI,KAAK,IAAI,CAAC,EAAE,CAAC;gBACtB,cAAc,CAAC,IAAI,CAAC;oBAClB,KAAK,EAAE,WAAW,GAAG,mBAAmB;oBACxC,GAAG,EAAE,KAAK,CAAC,GAAG,GAAG,IAAI;oBACrB,SAAS,EAAE,IAAI,GAAG,UAAU,GAAG,OAAO;iBACvC,CAAC,CAAC;YACL,CAAC;QACH,CAAC;QACD,MAAM,WAAW,GAAG,qBAAqB,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;QAC3D,IAAI,WAAW,EAAE,CAAC;YAChB,MAAM,SAAS,GAAG,MAAM,CAAC,KAAK,GAAG,CAAC,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC;YACnD,IAAI,MAAM,CAAC,QAAQ,KAAK,UAAU,IAAI,KAAK,CAAC,OAAO,EAAE,CAAC;gBACpD,KAAK,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,GAAG,EAAE,KAAK,CAAC,GAAG,EAAE,MAAM,EAAE,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,UAAU,CAAC,EAAE,CAAC,CAAC;YAC7F,CAAC;iBAAM,IAAI,MAAM,CAAC,QAAQ,KAAK,UAAU,EAAE,CAAC;gBAC1C,KAAK,CAAC,IAAI,CAAC;oBACT,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,GAAG,IAAI,CAAC;oBAC9C,GAAG,EAAE,KAAK,CAAC,GAAG;oBACd,MAAM,EAAE,CAAC,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,UAAU,CAAC;iBAC1C,CAAC,CAAC;YACL,CAAC;QACH,CAAC;QACD,WAAW,GAAG,WAAW,CAAC;IAC5B,CAAC;IACD,MAAM,KAAK,GAAkB;QAC3B,MAAM;QACN,KAAK,EAAE,cAAc;QACrB,IAAI,EAAE,aAAa;QACnB,KAAK,EAAE,cAAc;QACrB,OAAO,EAAE,cAAc;QACvB,OAAO,EAAE,cAAc;QACvB,KAAK;KACN,CAAC;IACF,MAAM,YAAY,GAAG,KAAK,CAAC;IAC3B,MAAM,IAAI,GAAG,KAAK,CAAC;IACnB,MAAM,KAAK,GAAiB,EAAE,CAAC;IAC/B,IAAI,KAAK,GAAG,QAAQ,CAAC;IACrB,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,IAAI,YAAY,GAAG,IAAI,EAAE,CAAC,IAAI,IAAI,EAAE,CAAC;QACpD,MAAM,EAAE,GAAG,eAAe,CAAC,kBAAkB,CAAC,KAAK,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,UAAU,CAAC,CAAC;QAC3E,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,GAAG,UAAU,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC;IACzD,CAAC;IACD,KAAK,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,EAAE,EAAE,CAAC;QAC9C,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,MAAM;YAAE,SAAS;QAC9C,IAAI,KAAK,GAAG,CAAC,CAAC;QACd,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,KAAK,EAAE,CAAC,EAAE;YAAE,KAAK,IAAI,MAAM,CAAC,CAAC,CAAC,EAAE,OAAO,IAAI,CAAC,CAAC;QACjE,MAAM,GAAG,GAAG,KAAK,GAAG,KAAK,CAAC,OAAO,CAAC;QAClC,KAAK,MAAM,KAAK,IAAI,KAAK,EAAE,CAAC;YAC1B,MAAM,CAAC,GAAG,KAAK,CAAC,MAAM,GAAG,UAAU,CAAC;YACpC,IAAI,CAAC,GAAG,KAAK,IAAI,CAAC,GAAG,GAAG;gBAAE,SAAS;YACnC,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC;YAClC,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;IACD,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QAC5B,KAAK,GAAG,MAAM,CAAC;QACf,KAAK,GAAG,MAAM,CAAC;IACjB,CAAC;IACD,IAAI,KAAK,GAAG,SAAS,IAAI,KAAK,GAAG,UAAU,GAAG,GAAG,IAAI,KAAK,GAAG,SAAS,EAAE,CAAC;QACvE,OAAO,CACL,KAAK,CAAC,UAAU,KAAK,SAAS,CAAC,CAAC,CAAC,kBAAkB,CAAC,CAAC,CAAC,SAAS,EAC/D,8BAA8B,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,gBAAgB,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,gCAAgC,SAAS,IAAI,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,UAAU,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,UAAU,UAAU,MAAM,EACrM;YACE,IAAI,EAAE,oBAAoB;YAC1B,IAAI,EAAE,0EAA0E;YAChF,aAAa,EAAE;gBACb,EAAE,MAAM,EAAE,aAAa,EAAE,WAAW,EAAE,iCAAiC,EAAE,IAAI,EAAE,kBAAkB,EAAE;gBACnG,EAAE,MAAM,EAAE,mBAAmB,EAAE,WAAW,EAAE,wCAAwC,EAAE,IAAI,EAAE,cAAc,EAAE;aAC7G;SACF,CACF,CAAC;IACJ,CAAC;IACD,IAAI,KAAK,GAAG,KAAK,CAAC,KAAK,IAAI,KAAK,GAAG,KAAK,CAAC,KAAK,EAAE,CAAC;QAC/C,QAAQ,CAAC,IAAI,CAAC;YACZ,QAAQ,EAAE,SAAS;YACnB,IAAI,EAAE,wBAAwB;YAC9B,OAAO,EAAE,qBAAqB,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,cAAc,KAAK,CAAC,EAAE,iBAAiB,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,KAAK,MAAM;YACzI,IAAI,EAAE,iDAAiD;YACvD,IAAI;SACL,CAAC,CAAC;IACL,CAAC;IACD,MAAM,QAAQ,GAAG,eAAe,CAAC,MAAM,EAAE,KAAK,EAAE,UAAU,CAAC,CAAC;IAC5D,MAAM,UAAU,GAAG,QAAQ,CAAC,UAAU,CAAC;IACvC,MAAM,QAAQ,GAAG,CAAC,OAAe,EAAU,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,GAAG,UAAU,CAAC,CAAC;IAC/E,MAAM,KAAK,GAAiB,SAAS,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC;QAC1D,KAAK;QACL,IAAI,EAAE,IAAI,CAAC,IAAI;QACf,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACtD,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;QAClD,WAAW,EAAE,IAAI,CAAC,WAAW;QAC7B,QAAQ,EAAE,IAAI,CAAC,QAAQ;KACxB,CAAC,CAAC,CAAC;IACJ,MAAM,OAAO,GAAmB,WAAW,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,EAAE;QAC/D,MAAM,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;QACtC,OAAO;YACL,KAAK;YACL,IAAI,EAAE,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC;YAC1E,IAAI,EAAE,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,gBAAgB,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,MAAM,CAAC;YAC/E,MAAM,EAAE,MAAM,EAAE,MAAM,IAAI,CAAC;YAC3B,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;YACvD,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,QAAQ,CAAC,MAAM,EAAE,QAAQ,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;YACnD,QAAQ,EAAE,MAAM,EAAE,QAAQ,IAAI,KAAK;SACpC,CAAC;IACJ,CAAC,CAAC,CAAC;IACH,KAAK,UAAU,CAAC;IAChB,OAAO;QACL,WAAW,EAAE,YAAY;QACzB,aAAa,EAAE,cAAc;QAC7B,IAAI,EAAE,QAAQ;QACd,UAAU;QACV,IAAI,EAAE,KAAK,CAAC,IAAI;QAChB,MAAM,EAAE,QAAQ,CAAC,MAAM;QACvB,KAAK;QACL,MAAM,EAAE,KAAK,CAAC,MAAM;QACpB,MAAM,EAAE,KAAK,CAAC,MAAM,IAAI,KAAK,CAAC,MAAM;QACpC,SAAS,EAAE,KAAK,CAAC,SAAS;QAC1B,OAAO,EAAE,KAAK,CAAC,OAAO;QACtB,QAAQ,EAAE,QAAQ,CAAC,QAAQ;QAC3B,KAAK;QACL,OAAO,EAAE,EAAE;QACX,KAAK;QACL,OAAO;QACP,KAAK,EAAE,EAAE;QACT,OAAO;QACP,QAAQ;QACR,WAAW,EAAE,EAAE;QACf,KAAK;QACL,KAAK;KACN,CAAC;AACJ,CAAC"}
|
|
@@ -0,0 +1,26 @@
|
|
|
1
|
+
import type { RenderStats } from "./renderer.ts";
|
|
2
|
+
import type { SynthesisPlan } from "./plan.ts";
|
|
3
|
+
export interface StreamOptions {
|
|
4
|
+
readonly blockFrames?: number;
|
|
5
|
+
readonly signal?: AbortSignal;
|
|
6
|
+
readonly yieldEvery?: number;
|
|
7
|
+
}
|
|
8
|
+
export interface PcmBlock {
|
|
9
|
+
readonly samples: Float32Array;
|
|
10
|
+
readonly start: number;
|
|
11
|
+
readonly sampleRate: number;
|
|
12
|
+
}
|
|
13
|
+
export type AudioSink = (block: Float32Array) => void | Promise<void>;
|
|
14
|
+
export declare function renderPcmStream(plan: SynthesisPlan, options?: StreamOptions): AsyncGenerator<PcmBlock, RenderStats, void>;
|
|
15
|
+
export declare function renderInto(plan: SynthesisPlan, sink: AudioSink, options?: StreamOptions): Promise<RenderStats>;
|
|
16
|
+
export declare function encodePlanAsync(plan: SynthesisPlan, options?: StreamOptions): Promise<{
|
|
17
|
+
readonly audio: Uint8Array;
|
|
18
|
+
readonly stats: RenderStats;
|
|
19
|
+
}>;
|
|
20
|
+
export declare function encodePlanSync(plan: SynthesisPlan, options?: StreamOptions): {
|
|
21
|
+
readonly audio: Uint8Array;
|
|
22
|
+
readonly stats: RenderStats;
|
|
23
|
+
};
|
|
24
|
+
export declare function pcm16Bytes(block: Float32Array): Uint8Array;
|
|
25
|
+
export declare function streamingWavHeader(sampleRate: number, channels?: number): Uint8Array;
|
|
26
|
+
//# sourceMappingURL=stream.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"stream.d.ts","sourceRoot":"","sources":["../../src/synth/stream.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAEjD,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,WAAW,CAAC;AAE/C,MAAM,WAAW,aAAa;IAC5B,QAAQ,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,MAAM,CAAC,EAAE,WAAW,CAAC;IAC9B,QAAQ,CAAC,UAAU,CAAC,EAAE,MAAM,CAAC;CAC9B;AAED,MAAM,WAAW,QAAQ;IACvB,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAC/B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;CAC7B;AAED,MAAM,MAAM,SAAS,GAAG,CAAC,KAAK,EAAE,YAAY,KAAK,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;AAEtE,wBAAuB,eAAe,CACpC,IAAI,EAAE,aAAa,EACnB,OAAO,GAAE,aAAkB,GAC1B,cAAc,CAAC,QAAQ,EAAE,WAAW,EAAE,IAAI,CAAC,CAkB7C;AAED,wBAAsB,UAAU,CAC9B,IAAI,EAAE,aAAa,EACnB,IAAI,EAAE,SAAS,EACf,OAAO,GAAE,aAAkB,GAC1B,OAAO,CAAC,WAAW,CAAC,CAQtB;AAED,wBAAsB,eAAe,CACnC,IAAI,EAAE,aAAa,EACnB,OAAO,GAAE,aAAkB,GAC1B,OAAO,CAAC;IAAE,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,WAAW,CAAA;CAAE,CAAC,CAatE;AAED,wBAAgB,cAAc,CAC5B,IAAI,EAAE,aAAa,EACnB,OAAO,GAAE,aAAkB,GAC1B;IAAE,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,WAAW,CAAA;CAAE,CAc7D;AAED,wBAAgB,UAAU,CAAC,KAAK,EAAE,YAAY,GAAG,UAAU,CAE1D;AAED,wBAAgB,kBAAkB,CAAC,UAAU,EAAE,MAAM,EAAE,QAAQ,GAAE,MAAU,GAAG,UAAU,CAMvF"}
|
|
@@ -0,0 +1,67 @@
|
|
|
1
|
+
import { setImmediate as yieldToEventLoop } from "node:timers/promises";
|
|
2
|
+
import { checkAbort } from "../errors.js";
|
|
3
|
+
import { encodePcm16, wavHeader } from "../wav.js";
|
|
4
|
+
import { PlanRenderer } from "./renderer.js";
|
|
5
|
+
export async function* renderPcmStream(plan, options = {}) {
|
|
6
|
+
const renderer = new PlanRenderer(plan);
|
|
7
|
+
const blockFrames = Math.max(64, Math.min(1 << 20, Math.round(options.blockFrames ?? Math.max(1024, Math.round(plan.sampleRate / 20)))));
|
|
8
|
+
const yieldEvery = Math.max(1, options.yieldEvery ?? 4);
|
|
9
|
+
let produced = 0;
|
|
10
|
+
while (!renderer.finished) {
|
|
11
|
+
checkAbort(options.signal);
|
|
12
|
+
const start = renderer.position;
|
|
13
|
+
const block = new Float32Array(Math.min(blockFrames, renderer.frames - start));
|
|
14
|
+
renderer.fill(block);
|
|
15
|
+
yield { samples: block, start, sampleRate: plan.sampleRate };
|
|
16
|
+
produced += 1;
|
|
17
|
+
if (produced % yieldEvery === 0)
|
|
18
|
+
await yieldToEventLoop();
|
|
19
|
+
}
|
|
20
|
+
return renderer.stats();
|
|
21
|
+
}
|
|
22
|
+
export async function renderInto(plan, sink, options = {}) {
|
|
23
|
+
const iterator = renderPcmStream(plan, options);
|
|
24
|
+
while (true) {
|
|
25
|
+
const next = await iterator.next();
|
|
26
|
+
if (next.done)
|
|
27
|
+
return next.value;
|
|
28
|
+
await sink(next.value.samples);
|
|
29
|
+
checkAbort(options.signal);
|
|
30
|
+
}
|
|
31
|
+
}
|
|
32
|
+
export async function encodePlanAsync(plan, options = {}) {
|
|
33
|
+
const out = new Uint8Array(44 + plan.frames * 2);
|
|
34
|
+
out.set(wavHeader(plan.frames, plan.sampleRate));
|
|
35
|
+
let offset = 44;
|
|
36
|
+
const stats = await renderInto(plan, (block) => {
|
|
37
|
+
encodePcm16(block, out, offset);
|
|
38
|
+
offset += block.length * 2;
|
|
39
|
+
}, options);
|
|
40
|
+
return { audio: out, stats };
|
|
41
|
+
}
|
|
42
|
+
export function encodePlanSync(plan, options = {}) {
|
|
43
|
+
const renderer = new PlanRenderer(plan);
|
|
44
|
+
const out = new Uint8Array(44 + plan.frames * 2);
|
|
45
|
+
out.set(wavHeader(plan.frames, plan.sampleRate));
|
|
46
|
+
const step = Math.max(1024, Math.round(plan.sampleRate / 4));
|
|
47
|
+
const buffer = new Float32Array(step);
|
|
48
|
+
let offset = 44;
|
|
49
|
+
while (!renderer.finished) {
|
|
50
|
+
checkAbort(options.signal);
|
|
51
|
+
const count = renderer.fill(buffer, 0, Math.min(step, renderer.frames - renderer.position));
|
|
52
|
+
encodePcm16(buffer.subarray(0, count), out, offset);
|
|
53
|
+
offset += count * 2;
|
|
54
|
+
}
|
|
55
|
+
return { audio: out, stats: renderer.stats() };
|
|
56
|
+
}
|
|
57
|
+
export function pcm16Bytes(block) {
|
|
58
|
+
return encodePcm16(block);
|
|
59
|
+
}
|
|
60
|
+
export function streamingWavHeader(sampleRate, channels = 1) {
|
|
61
|
+
const header = wavHeader(0, sampleRate, channels);
|
|
62
|
+
const view = new DataView(header.buffer);
|
|
63
|
+
view.setUint32(4, 0xffffffff, true);
|
|
64
|
+
view.setUint32(40, 0xffffffff, true);
|
|
65
|
+
return header;
|
|
66
|
+
}
|
|
67
|
+
//# sourceMappingURL=stream.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"stream.js","sourceRoot":"","sources":["../../src/synth/stream.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,IAAI,gBAAgB,EAAE,MAAM,sBAAsB,CAAC;AACxE,OAAO,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAC1C,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,WAAW,CAAC;AAEnD,OAAO,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAiB7C,MAAM,CAAC,KAAK,SAAS,CAAC,CAAC,eAAe,CACpC,IAAmB,EACnB,UAAyB,EAAE;IAE3B,MAAM,QAAQ,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,CAAC;IACxC,MAAM,WAAW,GAAG,IAAI,CAAC,GAAG,CAC1B,EAAE,EACF,IAAI,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,WAAW,IAAI,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CACvG,CAAC;IACF,MAAM,UAAU,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,CAAC,UAAU,IAAI,CAAC,CAAC,CAAC;IACxD,IAAI,QAAQ,GAAG,CAAC,CAAC;IACjB,OAAO,CAAC,QAAQ,CAAC,QAAQ,EAAE,CAAC;QAC1B,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QAC3B,MAAM,KAAK,GAAG,QAAQ,CAAC,QAAQ,CAAC;QAChC,MAAM,KAAK,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,GAAG,CAAC,WAAW,EAAE,QAAQ,CAAC,MAAM,GAAG,KAAK,CAAC,CAAC,CAAC;QAC/E,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACrB,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,EAAE,CAAC;QAC7D,QAAQ,IAAI,CAAC,CAAC;QACd,IAAI,QAAQ,GAAG,UAAU,KAAK,CAAC;YAAE,MAAM,gBAAgB,EAAE,CAAC;IAC5D,CAAC;IACD,OAAO,QAAQ,CAAC,KAAK,EAAE,CAAC;AAC1B,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,UAAU,CAC9B,IAAmB,EACnB,IAAe,EACf,UAAyB,EAAE;IAE3B,MAAM,QAAQ,GAAG,eAAe,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;IAChD,OAAO,IAAI,EAAE,CAAC;QACZ,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;QACnC,IAAI,IAAI,CAAC,IAAI;YAAE,OAAO,IAAI,CAAC,KAAK,CAAC;QACjC,MAAM,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;QAC/B,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;IAC7B,CAAC;AACH,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,eAAe,CACnC,IAAmB,EACnB,UAAyB,EAAE;IAE3B,MAAM,GAAG,GAAG,IAAI,UAAU,CAAC,EAAE,GAAG,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACjD,GAAG,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;IACjD,IAAI,MAAM,GAAG,EAAE,CAAC;IAChB,MAAM,KAAK,GAAG,MAAM,UAAU,CAC5B,IAAI,EACJ,CAAC,KAAK,EAAE,EAAE;QACR,WAAW,CAAC,KAAK,EAAE,GAAG,EAAE,MAAM,CAAC,CAAC;QAChC,MAAM,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC;IAC7B,CAAC,EACD,OAAO,CACR,CAAC;IACF,OAAO,EAAE,KAAK,EAAE,GAAG,EAAE,KAAK,EAAE,CAAC;AAC/B,CAAC;AAED,MAAM,UAAU,cAAc,CAC5B,IAAmB,EACnB,UAAyB,EAAE;IAE3B,MAAM,QAAQ,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,CAAC;IACxC,MAAM,GAAG,GAAG,IAAI,UAAU,CAAC,EAAE,GAAG,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IACjD,GAAG,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;IACjD,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,GAAG,CAAC,CAAC,CAAC,CAAC;IAC7D,MAAM,MAAM,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,CAAC;IACtC,IAAI,MAAM,GAAG,EAAE,CAAC;IAChB,OAAO,CAAC,QAAQ,CAAC,QAAQ,EAAE,CAAC;QAC1B,UAAU,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QAC3B,MAAM,KAAK,GAAG,QAAQ,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,QAAQ,CAAC,MAAM,GAAG,QAAQ,CAAC,QAAQ,CAAC,CAAC,CAAC;QAC5F,WAAW,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,EAAE,KAAK,CAAC,EAAE,GAAG,EAAE,MAAM,CAAC,CAAC;QACpD,MAAM,IAAI,KAAK,GAAG,CAAC,CAAC;IACtB,CAAC;IACD,OAAO,EAAE,KAAK,EAAE,GAAG,EAAE,KAAK,EAAE,QAAQ,CAAC,KAAK,EAAE,EAAE,CAAC;AACjD,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,KAAmB;IAC5C,OAAO,WAAW,CAAC,KAAK,CAAC,CAAC;AAC5B,CAAC;AAED,MAAM,UAAU,kBAAkB,CAAC,UAAkB,EAAE,WAAmB,CAAC;IACzE,MAAM,MAAM,GAAG,SAAS,CAAC,CAAC,EAAE,UAAU,EAAE,QAAQ,CAAC,CAAC;IAClD,MAAM,IAAI,GAAG,IAAI,QAAQ,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IACzC,IAAI,CAAC,SAAS,CAAC,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,CAAC;IACpC,IAAI,CAAC,SAAS,CAAC,EAAE,EAAE,UAAU,EAAE,IAAI,CAAC,CAAC;IACrC,OAAO,MAAM,CAAC;AAChB,CAAC"}
|
|
@@ -0,0 +1,35 @@
|
|
|
1
|
+
import type { Diagnostic } from "../errors.ts";
|
|
2
|
+
import type { ResolvedSpeech } from "../types.ts";
|
|
3
|
+
import type { CompileOptions } from "./engine.ts";
|
|
4
|
+
import type { PcmBlock } from "./stream.ts";
|
|
5
|
+
export interface TextStreamOptions extends CompileOptions {
|
|
6
|
+
readonly blockFrames?: number;
|
|
7
|
+
readonly flushMs?: number;
|
|
8
|
+
readonly maxSentenceChars?: number;
|
|
9
|
+
}
|
|
10
|
+
export type TextStreamEvent = {
|
|
11
|
+
readonly type: "sentence";
|
|
12
|
+
readonly index: number;
|
|
13
|
+
readonly text: string;
|
|
14
|
+
readonly kana: string;
|
|
15
|
+
readonly durationSeconds: number;
|
|
16
|
+
readonly frames: number;
|
|
17
|
+
readonly warnings: readonly Diagnostic[];
|
|
18
|
+
readonly startFrame: number;
|
|
19
|
+
} | {
|
|
20
|
+
readonly type: "audio";
|
|
21
|
+
readonly index: number;
|
|
22
|
+
readonly block: PcmBlock;
|
|
23
|
+
readonly sequence: number;
|
|
24
|
+
} | {
|
|
25
|
+
readonly type: "end";
|
|
26
|
+
readonly sentences: number;
|
|
27
|
+
readonly frames: number;
|
|
28
|
+
readonly durationSeconds: number;
|
|
29
|
+
};
|
|
30
|
+
export declare function takeSentences(buffer: string, maxChars: number): {
|
|
31
|
+
readonly sentences: string[];
|
|
32
|
+
readonly rest: string;
|
|
33
|
+
};
|
|
34
|
+
export declare function synthesizeTextStream(chunks: AsyncIterable<string>, request: ResolvedSpeech, options?: TextStreamOptions): AsyncGenerator<TextStreamEvent, void, void>;
|
|
35
|
+
//# sourceMappingURL=textstream.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"textstream.d.ts","sourceRoot":"","sources":["../../src/synth/textstream.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,cAAc,CAAC;AAI/C,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAClD,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAKlD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAG5C,MAAM,WAAW,iBAAkB,SAAQ,cAAc;IACvD,QAAQ,CAAC,WAAW,CAAC,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,OAAO,CAAC,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC;CACpC;AAED,MAAM,MAAM,eAAe,GACvB;IACE,QAAQ,CAAC,IAAI,EAAE,UAAU,CAAC;IAC1B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IACvB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;IACjC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,QAAQ,EAAE,SAAS,UAAU,EAAE,CAAC;IACzC,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;CAC7B,GACD;IAAE,QAAQ,CAAC,IAAI,EAAE,OAAO,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;IAAC,QAAQ,CAAC,KAAK,EAAE,QAAQ,CAAC;IAAC,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,GACvG;IAAE,QAAQ,CAAC,IAAI,EAAE,KAAK,CAAC;IAAC,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAAC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IAAC,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAA;CAAE,CAAC;AAIpH,wBAAgB,aAAa,CAC3B,MAAM,EAAE,MAAM,EACd,QAAQ,EAAE,MAAM,GACf;IAAE,QAAQ,CAAC,SAAS,EAAE,MAAM,EAAE,CAAC;IAAC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAA;CAAE,CAsBzD;AAUD,wBAAuB,oBAAoB,CACzC,MAAM,EAAE,aAAa,CAAC,MAAM,CAAC,EAC7B,OAAO,EAAE,cAAc,EACvB,OAAO,GAAE,iBAAsB,GAC9B,cAAc,CAAC,eAAe,EAAE,IAAI,EAAE,IAAI,CAAC,CAwF7C"}
|
|
@@ -0,0 +1,127 @@
|
|
|
1
|
+
import { checkAbort } from "../errors.js";
|
|
2
|
+
import { LIMITS } from "../limits.js";
|
|
3
|
+
import { compileRequest } from "./engine.js";
|
|
4
|
+
import { boundaryPauseSeconds, utteranceFinalPauseSeconds } from "./speech.js";
|
|
5
|
+
import { renderPcmStream } from "./stream.js";
|
|
6
|
+
const TERMINALS = /[。!?!?\n]/u;
|
|
7
|
+
export function takeSentences(buffer, maxChars) {
|
|
8
|
+
const sentences = [];
|
|
9
|
+
let rest = buffer;
|
|
10
|
+
while (rest.length > 0) {
|
|
11
|
+
const match = TERMINALS.exec(rest);
|
|
12
|
+
if (match === null) {
|
|
13
|
+
if (Array.from(rest).length >= maxChars) {
|
|
14
|
+
const chars = Array.from(rest);
|
|
15
|
+
const cut = lastPauseIndex(chars, maxChars);
|
|
16
|
+
sentences.push(chars.slice(0, cut).join(""));
|
|
17
|
+
rest = chars.slice(cut).join("");
|
|
18
|
+
continue;
|
|
19
|
+
}
|
|
20
|
+
break;
|
|
21
|
+
}
|
|
22
|
+
let end = match.index + match[0].length;
|
|
23
|
+
while (end < rest.length && TERMINALS.test(rest[end] ?? ""))
|
|
24
|
+
end += 1;
|
|
25
|
+
const sentence = rest.slice(0, end);
|
|
26
|
+
rest = rest.slice(end);
|
|
27
|
+
if (sentence.trim().length > 0)
|
|
28
|
+
sentences.push(sentence);
|
|
29
|
+
}
|
|
30
|
+
return { sentences, rest };
|
|
31
|
+
}
|
|
32
|
+
function lastPauseIndex(chars, maxChars) {
|
|
33
|
+
for (let index = Math.min(chars.length, maxChars) - 1; index > maxChars / 2; index--) {
|
|
34
|
+
const char = chars[index];
|
|
35
|
+
if (char === "、" || char === "," || char === " " || char === " ")
|
|
36
|
+
return index + 1;
|
|
37
|
+
}
|
|
38
|
+
return Math.min(chars.length, maxChars);
|
|
39
|
+
}
|
|
40
|
+
export async function* synthesizeTextStream(chunks, request, options = {}) {
|
|
41
|
+
const maxChars = options.maxSentenceChars ?? LIMITS.streamSentenceChars;
|
|
42
|
+
const sampleRate = options.sampleRate ?? 24_000;
|
|
43
|
+
const timing = {
|
|
44
|
+
speed: request.speed,
|
|
45
|
+
pauseLength: request.pauseLength,
|
|
46
|
+
pauseScale: request.pauseScale,
|
|
47
|
+
};
|
|
48
|
+
let buffer = "";
|
|
49
|
+
let index = 0;
|
|
50
|
+
let sequence = 0;
|
|
51
|
+
let frames = 0;
|
|
52
|
+
let lastBoundary;
|
|
53
|
+
let lastParagraphEnd = false;
|
|
54
|
+
let pendingPlan;
|
|
55
|
+
const leadPause = () => {
|
|
56
|
+
if (lastBoundary === undefined)
|
|
57
|
+
return request.prePause;
|
|
58
|
+
return Math.max(0, boundaryPauseSeconds(lastBoundary, timing, lastParagraphEnd) - utteranceFinalPauseSeconds(lastBoundary, timing));
|
|
59
|
+
};
|
|
60
|
+
const compile = (text) => compileRequest({ ...request, text, prePause: leadPause(), postPause: 0 }, sampleRate, options).then((compiled) => compiled.plan);
|
|
61
|
+
const emit = async function* (text) {
|
|
62
|
+
const plan = await (pendingPlan ?? compile(text));
|
|
63
|
+
pendingPlan = undefined;
|
|
64
|
+
lastBoundary = plan.reading?.phrases.at(-1)?.boundary ?? "sentence";
|
|
65
|
+
lastParagraphEnd = plan.reading?.sentences.at(-1)?.paragraphEnd ?? false;
|
|
66
|
+
const startFrame = frames;
|
|
67
|
+
yield {
|
|
68
|
+
type: "sentence",
|
|
69
|
+
index,
|
|
70
|
+
text,
|
|
71
|
+
kana: plan.reading?.kana ?? "",
|
|
72
|
+
durationSeconds: plan.frames / plan.sampleRate,
|
|
73
|
+
frames: plan.frames,
|
|
74
|
+
warnings: plan.warnings,
|
|
75
|
+
startFrame,
|
|
76
|
+
};
|
|
77
|
+
const iterator = renderPcmStream(plan, {
|
|
78
|
+
...(options.blockFrames === undefined ? {} : { blockFrames: options.blockFrames }),
|
|
79
|
+
...(options.signal === undefined ? {} : { signal: options.signal }),
|
|
80
|
+
});
|
|
81
|
+
while (true) {
|
|
82
|
+
const next = await iterator.next();
|
|
83
|
+
if (next.done)
|
|
84
|
+
break;
|
|
85
|
+
yield { type: "audio", index, block: { ...next.value, start: next.value.start + startFrame }, sequence };
|
|
86
|
+
sequence += 1;
|
|
87
|
+
}
|
|
88
|
+
frames += plan.frames;
|
|
89
|
+
index += 1;
|
|
90
|
+
};
|
|
91
|
+
const queue = [];
|
|
92
|
+
for await (const chunk of chunks) {
|
|
93
|
+
checkAbort(options.signal);
|
|
94
|
+
buffer += chunk;
|
|
95
|
+
const taken = takeSentences(buffer, maxChars);
|
|
96
|
+
buffer = taken.rest;
|
|
97
|
+
queue.push(...taken.sentences);
|
|
98
|
+
while (queue.length > 0) {
|
|
99
|
+
const text = queue.shift();
|
|
100
|
+
if (text === undefined)
|
|
101
|
+
break;
|
|
102
|
+
const following = queue[0];
|
|
103
|
+
yield* emit(text);
|
|
104
|
+
if (following !== undefined)
|
|
105
|
+
pendingPlan = compile(following);
|
|
106
|
+
}
|
|
107
|
+
}
|
|
108
|
+
const tail = buffer.trim();
|
|
109
|
+
if (tail.length > 0) {
|
|
110
|
+
pendingPlan = undefined;
|
|
111
|
+
yield* emit(tail);
|
|
112
|
+
}
|
|
113
|
+
const trailingFrames = index === 0 ? 0 : Math.round(request.postPause * sampleRate);
|
|
114
|
+
if (trailingFrames > 0) {
|
|
115
|
+
checkAbort(options.signal);
|
|
116
|
+
yield {
|
|
117
|
+
type: "audio",
|
|
118
|
+
index: index - 1,
|
|
119
|
+
block: { samples: new Float32Array(trailingFrames), start: frames, sampleRate },
|
|
120
|
+
sequence,
|
|
121
|
+
};
|
|
122
|
+
sequence += 1;
|
|
123
|
+
frames += trailingFrames;
|
|
124
|
+
}
|
|
125
|
+
yield { type: "end", sentences: index, frames, durationSeconds: frames / sampleRate };
|
|
126
|
+
}
|
|
127
|
+
//# sourceMappingURL=textstream.js.map
|