Add JavaScript API (WebAssembly) for Kokoro TTS 1.0 (#1809)

.github/scripts/test-nodejs-npm.sh

@@ -10,12 +10,21 @@ ls -lh
 ls -lh node_modules
 
 # offline tts
+#
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-multi-lang-v1_0.tar.bz2
+tar xf kokoro-multi-lang-v1_0.tar.bz2
+rm kokoro-multi-lang-v1_0.tar.bz2
+
+node ./test-offline-tts-kokoro-zh-en.js
+ls -lh *.wav
+rm -rf kokoro-multi-lang-v1_0
 
 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/kokoro-en-v0_19.tar.bz2
 tar xf kokoro-en-v0_19.tar.bz2
 rm kokoro-en-v0_19.tar.bz2
 
 node ./test-offline-tts-kokoro-en.js
+rm -rf kokoro-en-v0_19
 
 ls -lh
 

nodejs-examples/test-offline-tts-kokoro-zh-en.js

@@ -0,0 +1,40 @@
+// Copyright (c)  2025  Xiaomi Corporation (authors: Fangjun Kuang)
+
+const sherpa_onnx = require('sherpa-onnx');
+
+function createOfflineTts() {
+  let offlineTtsKokoroModelConfig = {
+    model: './kokoro-multi-lang-v1_0/model.onnx',
+    voices: './kokoro-multi-lang-v1_0/voices.bin',
+    tokens: './kokoro-multi-lang-v1_0/tokens.txt',
+    dataDir: './kokoro-multi-lang-v1_0/espeak-ng-data',
+    dictDir: './kokoro-multi-lang-v1_0/dict',
+    lexicon:
+        './kokoro-multi-lang-v1_0/lexicon-us-en.txt,./kokoro-multi-lang-v1_0/lexicon-zh.txt',
+    lengthScale: 1.0,
+  };
+  let offlineTtsModelConfig = {
+    offlineTtsKokoroModelConfig: offlineTtsKokoroModelConfig,
+    numThreads: 1,
+    debug: 1,
+    provider: 'cpu',
+  };
+
+  let offlineTtsConfig = {
+    offlineTtsModelConfig: offlineTtsModelConfig,
+    maxNumSentences: 1,
+  };
+
+  return sherpa_onnx.createOfflineTts(offlineTtsConfig);
+}
+
+const tts = createOfflineTts();
+const speakerId = 49;
+const speed = 1.0;
+const text =
+    '中英文语音合成测试。This is generated by next generation Kaldi using Kokoro without Misaki. 你觉得中英文说的如何呢？'
+
+const audio = tts.generate({text: text, sid: speakerId, speed: speed});
+tts.save('./test-kokoro-zh-en-49.wav', audio);
+console.log('Saved to test-kokoro-zh-en-49.wav successfully.');
+tts.free();

wasm/tts/sherpa-onnx-tts.js

@@ -141,12 +141,15 @@ function initSherpaOnnxOfflineTtsKokoroModelConfig(config, Module) {
   const voicesLen = Module.lengthBytesUTF8(config.voices) + 1;
   const tokensLen = Module.lengthBytesUTF8(config.tokens || '') + 1;
   const dataDirLen = Module.lengthBytesUTF8(config.dataDir || '') + 1;
+  const dictDirLen = Module.lengthBytesUTF8(config.dictDir || '') + 1;
+  const lexiconLen = Module.lengthBytesUTF8(config.lexicon || '') + 1;
 
-  const n = modelLen + voicesLen + tokensLen + dataDirLen;
+  const n =
+      modelLen + voicesLen + tokensLen + dataDirLen + dictDirLen + lexiconLen;
 
   const buffer = Module._malloc(n);
 
-  const len = 5 * 4;
+  const len = 7 * 4;
   const ptr = Module._malloc(len);
 
   let offset = 0;
@@ -162,6 +165,12 @@ function initSherpaOnnxOfflineTtsKokoroModelConfig(config, Module) {
   Module.stringToUTF8(config.dataDir || '', buffer + offset, dataDirLen);
   offset += dataDirLen;
 
+  Module.stringToUTF8(config.dictDir || '', buffer + offset, dictDirLen);
+  offset += dictDirLen;
+
+  Module.stringToUTF8(config.lexicon || '', buffer + offset, lexiconLen);
+  offset += lexiconLen;
+
   offset = 0;
   Module.setValue(ptr, buffer + offset, 'i8*');
   offset += modelLen;
@@ -177,6 +186,12 @@ function initSherpaOnnxOfflineTtsKokoroModelConfig(config, Module) {
 
   Module.setValue(ptr + 16, config.lengthScale || 1.0, 'float');
 
+  Module.setValue(ptr + 20, buffer + offset, 'i8*');
+  offset += dictDirLen;
+
+  Module.setValue(ptr + 24, buffer + offset, 'i8*');
+  offset += lexiconLen;
+
   return {
     buffer: buffer, ptr: ptr, len: len,
   }
@@ -216,6 +231,8 @@ function initSherpaOnnxOfflineTtsModelConfig(config, Module) {
       tokens: '',
       lengthScale: 1.0,
       dataDir: '',
+      dictDir: '',
+      lexicon: '',
     };
   }
 
@@ -382,6 +399,8 @@ function createOfflineTts(Module, myConfig) {
     tokens: '',
     dataDir: '',
     lengthScale: 1.0,
+    dictDir: '',
+    lexicon: '',
   };
 
   const offlineTtsModelConfig = {

wasm/tts/sherpa-onnx-wasm-main-tts.cc

@@ -15,7 +15,7 @@ extern "C" {
 
 static_assert(sizeof(SherpaOnnxOfflineTtsVitsModelConfig) == 8 * 4, "");
 static_assert(sizeof(SherpaOnnxOfflineTtsMatchaModelConfig) == 8 * 4, "");
-static_assert(sizeof(SherpaOnnxOfflineTtsKokoroModelConfig) == 5 * 4, "");
+static_assert(sizeof(SherpaOnnxOfflineTtsKokoroModelConfig) == 7 * 4, "");
 static_assert(sizeof(SherpaOnnxOfflineTtsModelConfig) ==
                   sizeof(SherpaOnnxOfflineTtsVitsModelConfig) +
                       sizeof(SherpaOnnxOfflineTtsMatchaModelConfig) +
@@ -56,6 +56,8 @@ void MyPrint(SherpaOnnxOfflineTtsConfig *tts_config) {
   fprintf(stdout, "tokens: %s\n", kokoro->tokens);
   fprintf(stdout, "data_dir: %s\n", kokoro->data_dir);
   fprintf(stdout, "length scale: %.3f\n", kokoro->length_scale);
+  fprintf(stdout, "dict_dir: %s\n", kokoro->dict_dir);
+  fprintf(stdout, "lexicon: %s\n", kokoro->lexicon);
 
   fprintf(stdout, "----------tts model config----------\n");
   fprintf(stdout, "num threads: %d\n", tts_model_config->num_threads);