Fangjun Kuang
Committed by GitHub

Add JavaScript API (node-addon) for Kokoro TTS models (#1725)

@@ -85,6 +85,13 @@ fi @@ -85,6 +85,13 @@ fi
85 85
86 echo "----------tts----------" 86 echo "----------tts----------"
87 87
  88 +curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-en-v0_19.tar.bz2
  89 +tar xf kokoro-en-v0_19.tar.bz2
  90 +rm kokoro-en-v0_19.tar.bz2
  91 +
  92 +node ./test_tts_non_streaming_kokoro_en.js
  93 +ls -lh *.wav
  94 +
88 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/matcha-icefall-en_US-ljspeech.tar.bz2 95 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/matcha-icefall-en_US-ljspeech.tar.bz2
89 tar xvf matcha-icefall-en_US-ljspeech.tar.bz2 96 tar xvf matcha-icefall-en_US-ljspeech.tar.bz2
90 rm matcha-icefall-en_US-ljspeech.tar.bz2 97 rm matcha-icefall-en_US-ljspeech.tar.bz2
@@ -53,6 +53,25 @@ static SherpaOnnxOfflineTtsMatchaModelConfig GetOfflineTtsMatchaModelConfig( @@ -53,6 +53,25 @@ static SherpaOnnxOfflineTtsMatchaModelConfig GetOfflineTtsMatchaModelConfig(
53 return c; 53 return c;
54 } 54 }
55 55
  56 +static SherpaOnnxOfflineTtsKokoroModelConfig GetOfflineTtsKokoroModelConfig(
  57 + Napi::Object obj) {
  58 + SherpaOnnxOfflineTtsKokoroModelConfig c;
  59 + memset(&c, 0, sizeof(c));
  60 +
  61 + if (!obj.Has("kokoro") || !obj.Get("kokoro").IsObject()) {
  62 + return c;
  63 + }
  64 +
  65 + Napi::Object o = obj.Get("kokoro").As<Napi::Object>();
  66 + SHERPA_ONNX_ASSIGN_ATTR_STR(model, model);
  67 + SHERPA_ONNX_ASSIGN_ATTR_STR(voices, voices);
  68 + SHERPA_ONNX_ASSIGN_ATTR_STR(tokens, tokens);
  69 + SHERPA_ONNX_ASSIGN_ATTR_STR(data_dir, dataDir);
  70 + SHERPA_ONNX_ASSIGN_ATTR_FLOAT(length_scale, lengthScale);
  71 +
  72 + return c;
  73 +}
  74 +
56 static SherpaOnnxOfflineTtsModelConfig GetOfflineTtsModelConfig( 75 static SherpaOnnxOfflineTtsModelConfig GetOfflineTtsModelConfig(
57 Napi::Object obj) { 76 Napi::Object obj) {
58 SherpaOnnxOfflineTtsModelConfig c; 77 SherpaOnnxOfflineTtsModelConfig c;
@@ -66,6 +85,7 @@ static SherpaOnnxOfflineTtsModelConfig GetOfflineTtsModelConfig( @@ -66,6 +85,7 @@ static SherpaOnnxOfflineTtsModelConfig GetOfflineTtsModelConfig(
66 85
67 c.vits = GetOfflineTtsVitsModelConfig(o); 86 c.vits = GetOfflineTtsVitsModelConfig(o);
68 c.matcha = GetOfflineTtsMatchaModelConfig(o); 87 c.matcha = GetOfflineTtsMatchaModelConfig(o);
  88 + c.kokoro = GetOfflineTtsKokoroModelConfig(o);
69 89
70 SHERPA_ONNX_ASSIGN_ATTR_INT32(num_threads, numThreads); 90 SHERPA_ONNX_ASSIGN_ATTR_INT32(num_threads, numThreads);
71 91
@@ -180,6 +200,22 @@ static Napi::External<SherpaOnnxOfflineTts> CreateOfflineTtsWrapper( @@ -180,6 +200,22 @@ static Napi::External<SherpaOnnxOfflineTts> CreateOfflineTtsWrapper(
180 delete[] c.model.matcha.dict_dir; 200 delete[] c.model.matcha.dict_dir;
181 } 201 }
182 202
  203 + if (c.model.kokoro.model) {
  204 + delete[] c.model.kokoro.model;
  205 + }
  206 +
  207 + if (c.model.kokoro.voices) {
  208 + delete[] c.model.kokoro.voices;
  209 + }
  210 +
  211 + if (c.model.kokoro.tokens) {
  212 + delete[] c.model.kokoro.tokens;
  213 + }
  214 +
  215 + if (c.model.kokoro.data_dir) {
  216 + delete[] c.model.kokoro.data_dir;
  217 + }
  218 +
183 if (c.model.provider) { 219 if (c.model.provider) {
184 delete[] c.model.provider; 220 delete[] c.model.provider;
185 } 221 }
@@ -133,6 +133,7 @@ The following tables list the examples in this folder. @@ -133,6 +133,7 @@ The following tables list the examples in this folder.
133 133
134 |File| Description| 134 |File| Description|
135 |---|---| 135 |---|---|
  136 +|[./test_tts_non_streaming_kokoro_en.js](./test_tts_non_streaming_kokoro_en.js)| Text-to-speech with a Kokoro English Model|
136 |[./test_tts_non_streaming_matcha_icefall_en.js](./test_tts_non_streaming_matcha_icefall_en.js)| Text-to-speech with a [MatchaTTS English Model](https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/matcha.html#matcha-icefall-en-us-ljspeech-american-english-1-female-speaker)| 137 |[./test_tts_non_streaming_matcha_icefall_en.js](./test_tts_non_streaming_matcha_icefall_en.js)| Text-to-speech with a [MatchaTTS English Model](https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/matcha.html#matcha-icefall-en-us-ljspeech-american-english-1-female-speaker)|
137 |[./test_tts_non_streaming_matcha_icefall_zhjs](./test_tts_non_streaming_matcha_icefall_zh.js)| Text-to-speech with a [MatchaTTS Chinese Model](https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/matcha.html#matcha-icefall-zh-baker-chinese-1-female-speaker)| 138 |[./test_tts_non_streaming_matcha_icefall_zhjs](./test_tts_non_streaming_matcha_icefall_zh.js)| Text-to-speech with a [MatchaTTS Chinese Model](https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/matcha.html#matcha-icefall-zh-baker-chinese-1-female-speaker)|
138 |[./test_tts_non_streaming_vits_piper_en.js](./test_tts_non_streaming_vits_piper_en.js)| Text-to-speech with a [piper](https://github.com/rhasspy/piper) English model| 139 |[./test_tts_non_streaming_vits_piper_en.js](./test_tts_non_streaming_vits_piper_en.js)| Text-to-speech with a [piper](https://github.com/rhasspy/piper) English model|
@@ -347,6 +348,16 @@ npm install naudiodon2 @@ -347,6 +348,16 @@ npm install naudiodon2
347 node ./test_vad_asr_non_streaming_sense_voice_microphone.js 348 node ./test_vad_asr_non_streaming_sense_voice_microphone.js
348 ``` 349 ```
349 350
  351 +### Text-to-speech with Kokoro TTS models (English TTS)
  352 +
  353 +```bash
  354 +curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-en-v0_19.tar.bz2
  355 +tar xf kokoro-en-v0_19.tar.bz2
  356 +rm kokoro-en-v0_19.tar.bz2
  357 +
  358 +node ./test_tts_non_streaming_kokoro_en.js
  359 +```
  360 +
350 ### Text-to-speech with MatchaTTS models (English TTS) 361 ### Text-to-speech with MatchaTTS models (English TTS)
351 ```bash 362 ```bash
352 wget https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/matcha-icefall-en_US-ljspeech.tar.bz2 363 wget https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/matcha-icefall-en_US-ljspeech.tar.bz2
  1 +// Copyright (c) 2025 Xiaomi Corporation
  2 +const sherpa_onnx = require('sherpa-onnx-node');
  3 +
  4 +// please refer to
  5 +// https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/kokoro.html
  6 +// to download model files
  7 +function createOfflineTts() {
  8 + const config = {
  9 + model: {
  10 + kokoro: {
  11 + model: './kokoro-en-v0_19/model.onnx',
  12 + voices: './kokoro-en-v0_19/voices.bin',
  13 + tokens: './kokoro-en-v0_19/tokens.txt',
  14 + dataDir: './kokoro-en-v0_19/espeak-ng-data',
  15 + },
  16 + debug: true,
  17 + numThreads: 1,
  18 + provider: 'cpu',
  19 + },
  20 + maxNumSentences: 1,
  21 + };
  22 + return new sherpa_onnx.OfflineTts(config);
  23 +}
  24 +
  25 +const tts = createOfflineTts();
  26 +
  27 +const text =
  28 + 'Today as always, men fall into two groups: slaves and free men. Whoever does not have two-thirds of his day for himself, is a slave, whatever he may be: a statesman, a businessman, an official, or a scholar.'
  29 +
  30 +
  31 +let start = Date.now();
  32 +const audio = tts.generate({text: text, sid: 6, speed: 1.0});
  33 +let stop = Date.now();
  34 +const elapsed_seconds = (stop - start) / 1000;
  35 +const duration = audio.samples.length / audio.sampleRate;
  36 +const real_time_factor = elapsed_seconds / duration;
  37 +console.log('Wave duration', duration.toFixed(3), 'secodns')
  38 +console.log('Elapsed', elapsed_seconds.toFixed(3), 'secodns')
  39 +console.log(
  40 + `RTF = ${elapsed_seconds.toFixed(3)}/${duration.toFixed(3)} =`,
  41 + real_time_factor.toFixed(3))
  42 +
  43 +const filename = 'test-kokoro-en-6.wav';
  44 +sherpa_onnx.writeWave(
  45 + filename, {samples: audio.samples, sampleRate: audio.sampleRate});
  46 +
  47 +console.log(`Saved to ${filename}`);