Commit
·
b1d656c
1
Parent(s):
2a1d143
update model
Browse files- app-tts.js +1 -1
- sherpa-onnx-tts.js +96 -68
- sherpa-onnx-wasm-main-tts.js +0 -0
app-tts.js
CHANGED
|
@@ -22,7 +22,7 @@ Module.onRuntimeInitialized = function() {
|
|
| 22 |
console.log('Model files downloaded!');
|
| 23 |
|
| 24 |
console.log('Initializing tts ......');
|
| 25 |
-
tts =
|
| 26 |
if (tts.numSpeakers > 1) {
|
| 27 |
speakerIdLabel.innerHTML = `Speaker ID (0 - ${tts.numSpeakers - 1}):`;
|
| 28 |
}
|
|
|
|
| 22 |
console.log('Model files downloaded!');
|
| 23 |
|
| 24 |
console.log('Initializing tts ......');
|
| 25 |
+
tts = createOfflineTts(Module)
|
| 26 |
if (tts.numSpeakers > 1) {
|
| 27 |
speakerIdLabel.innerHTML = `Speaker ID (0 - ${tts.numSpeakers - 1}):`;
|
| 28 |
}
|
sherpa-onnx-tts.js
CHANGED
|
@@ -1,109 +1,109 @@
|
|
| 1 |
|
| 2 |
-
function freeConfig(config) {
|
| 3 |
if ('buffer' in config) {
|
| 4 |
-
_free(config.buffer);
|
| 5 |
}
|
| 6 |
|
| 7 |
if ('config' in config) {
|
| 8 |
-
freeConfig(config.config)
|
| 9 |
}
|
| 10 |
|
| 11 |
-
_free(config.ptr);
|
| 12 |
}
|
| 13 |
|
| 14 |
// The user should free the returned pointers
|
| 15 |
-
function initSherpaOnnxOfflineTtsVitsModelConfig(config) {
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
|
| 21 |
-
|
| 22 |
|
| 23 |
-
|
| 24 |
|
| 25 |
-
|
| 26 |
-
|
| 27 |
|
| 28 |
let offset = 0;
|
| 29 |
-
stringToUTF8(config.model, buffer + offset, modelLen);
|
| 30 |
offset += modelLen;
|
| 31 |
|
| 32 |
-
stringToUTF8(config.lexicon, buffer + offset, lexiconLen);
|
| 33 |
offset += lexiconLen;
|
| 34 |
|
| 35 |
-
stringToUTF8(config.tokens, buffer + offset, tokensLen);
|
| 36 |
offset += tokensLen;
|
| 37 |
|
| 38 |
-
stringToUTF8(config.dataDir, buffer + offset, dataDirLen);
|
| 39 |
offset += dataDirLen;
|
| 40 |
|
| 41 |
offset = 0;
|
| 42 |
-
setValue(ptr, buffer + offset, 'i8*');
|
| 43 |
offset += modelLen;
|
| 44 |
|
| 45 |
-
setValue(ptr + 4, buffer + offset, 'i8*');
|
| 46 |
offset += lexiconLen;
|
| 47 |
|
| 48 |
-
setValue(ptr + 8, buffer + offset, 'i8*');
|
| 49 |
offset += tokensLen;
|
| 50 |
|
| 51 |
-
setValue(ptr + 12, buffer + offset, 'i8*');
|
| 52 |
offset += dataDirLen;
|
| 53 |
|
| 54 |
-
setValue(ptr + 16, config.noiseScale, 'float');
|
| 55 |
-
setValue(ptr + 20, config.noiseScaleW, 'float');
|
| 56 |
-
setValue(ptr + 24, config.lengthScale, 'float');
|
| 57 |
|
| 58 |
return {
|
| 59 |
buffer: buffer, ptr: ptr, len: len,
|
| 60 |
}
|
| 61 |
}
|
| 62 |
|
| 63 |
-
function initSherpaOnnxOfflineTtsModelConfig(config) {
|
| 64 |
-
|
| 65 |
-
|
| 66 |
|
| 67 |
-
|
| 68 |
-
|
| 69 |
|
| 70 |
let offset = 0;
|
| 71 |
-
_CopyHeap(vitsModelConfig.ptr, vitsModelConfig.len, ptr + offset);
|
| 72 |
offset += vitsModelConfig.len;
|
| 73 |
|
| 74 |
-
setValue(ptr + offset, config.numThreads, 'i32');
|
| 75 |
offset += 4;
|
| 76 |
|
| 77 |
-
setValue(ptr + offset, config.debug, 'i32');
|
| 78 |
offset += 4;
|
| 79 |
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
stringToUTF8(config.provider, buffer, providerLen);
|
| 83 |
-
setValue(ptr + offset, buffer, 'i8*');
|
| 84 |
|
| 85 |
return {
|
| 86 |
buffer: buffer, ptr: ptr, len: len, config: vitsModelConfig,
|
| 87 |
}
|
| 88 |
}
|
| 89 |
|
| 90 |
-
function initSherpaOnnxOfflineTtsConfig(config) {
|
| 91 |
-
|
| 92 |
-
initSherpaOnnxOfflineTtsModelConfig(config.offlineTtsModelConfig);
|
| 93 |
-
|
| 94 |
-
|
| 95 |
|
| 96 |
let offset = 0;
|
| 97 |
-
_CopyHeap(modelConfig.ptr, modelConfig.len, ptr + offset);
|
| 98 |
offset += modelConfig.len;
|
| 99 |
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
stringToUTF8(config.ruleFsts, buffer, ruleFstsLen);
|
| 103 |
-
setValue(ptr + offset, buffer, 'i8*');
|
| 104 |
offset += 4;
|
| 105 |
|
| 106 |
-
setValue(ptr + offset, config.maxNumSentences, 'i32');
|
| 107 |
|
| 108 |
return {
|
| 109 |
buffer: buffer, ptr: ptr, len: len, config: modelConfig,
|
|
@@ -111,19 +111,21 @@ function initSherpaOnnxOfflineTtsConfig(config) {
|
|
| 111 |
}
|
| 112 |
|
| 113 |
class OfflineTts {
|
| 114 |
-
constructor(configObj) {
|
| 115 |
-
|
| 116 |
-
|
|
|
|
| 117 |
|
| 118 |
-
freeConfig(config);
|
| 119 |
|
| 120 |
this.handle = handle;
|
| 121 |
-
this.sampleRate = _SherpaOnnxOfflineTtsSampleRate(this.handle);
|
| 122 |
-
this.numSpeakers = _SherpaOnnxOfflineTtsNumSpeakers(this.handle);
|
|
|
|
| 123 |
}
|
| 124 |
|
| 125 |
free() {
|
| 126 |
-
_SherpaOnnxDestroyOfflineTts(this.handle);
|
| 127 |
this.handle = 0
|
| 128 |
}
|
| 129 |
|
|
@@ -133,29 +135,44 @@ class OfflineTts {
|
|
| 133 |
// speed: 1.0
|
| 134 |
// }
|
| 135 |
generate(config) {
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
stringToUTF8(config.text, textPtr, textLen);
|
| 139 |
|
| 140 |
-
|
| 141 |
this.handle, textPtr, config.sid, config.speed);
|
| 142 |
|
| 143 |
-
|
| 144 |
-
|
| 145 |
|
| 146 |
-
|
| 147 |
-
|
| 148 |
for (let i = 0; i < numSamples; i++) {
|
| 149 |
-
samples[i] = HEAPF32[samplesPtr + i];
|
| 150 |
}
|
| 151 |
|
| 152 |
-
_SherpaOnnxDestroyOfflineTtsGeneratedAudio(h);
|
| 153 |
return {samples: samples, sampleRate: sampleRate};
|
| 154 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 155 |
}
|
| 156 |
|
| 157 |
-
function
|
| 158 |
-
|
| 159 |
model: './model.onnx',
|
| 160 |
lexicon: '',
|
| 161 |
tokens: './tokens.txt',
|
|
@@ -164,7 +181,7 @@ function initSherpaOnnxOfflineTts() {
|
|
| 164 |
noiseScaleW: 0.8,
|
| 165 |
lengthScale: 1.0,
|
| 166 |
};
|
| 167 |
-
|
| 168 |
offlineTtsVitsModelConfig: offlineTtsVitsModelConfig,
|
| 169 |
numThreads: 1,
|
| 170 |
debug: 1,
|
|
@@ -176,5 +193,16 @@ function initSherpaOnnxOfflineTts() {
|
|
| 176 |
maxNumSentences: 1,
|
| 177 |
}
|
| 178 |
|
| 179 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 180 |
}
|
|
|
|
| 1 |
|
| 2 |
+
function freeConfig(config, Module) {
|
| 3 |
if ('buffer' in config) {
|
| 4 |
+
Module._free(config.buffer);
|
| 5 |
}
|
| 6 |
|
| 7 |
if ('config' in config) {
|
| 8 |
+
freeConfig(config.config, Module)
|
| 9 |
}
|
| 10 |
|
| 11 |
+
Module._free(config.ptr);
|
| 12 |
}
|
| 13 |
|
| 14 |
// The user should free the returned pointers
|
| 15 |
+
function initSherpaOnnxOfflineTtsVitsModelConfig(config, Module) {
|
| 16 |
+
const modelLen = Module.lengthBytesUTF8(config.model) + 1;
|
| 17 |
+
const lexiconLen = Module.lengthBytesUTF8(config.lexicon) + 1;
|
| 18 |
+
const tokensLen = Module.lengthBytesUTF8(config.tokens) + 1;
|
| 19 |
+
const dataDirLen = Module.lengthBytesUTF8(config.dataDir) + 1;
|
| 20 |
|
| 21 |
+
const n = modelLen + lexiconLen + tokensLen + dataDirLen;
|
| 22 |
|
| 23 |
+
const buffer = Module._malloc(n);
|
| 24 |
|
| 25 |
+
const len = 7 * 4;
|
| 26 |
+
const ptr = Module._malloc(len);
|
| 27 |
|
| 28 |
let offset = 0;
|
| 29 |
+
Module.stringToUTF8(config.model, buffer + offset, modelLen);
|
| 30 |
offset += modelLen;
|
| 31 |
|
| 32 |
+
Module.stringToUTF8(config.lexicon, buffer + offset, lexiconLen);
|
| 33 |
offset += lexiconLen;
|
| 34 |
|
| 35 |
+
Module.stringToUTF8(config.tokens, buffer + offset, tokensLen);
|
| 36 |
offset += tokensLen;
|
| 37 |
|
| 38 |
+
Module.stringToUTF8(config.dataDir, buffer + offset, dataDirLen);
|
| 39 |
offset += dataDirLen;
|
| 40 |
|
| 41 |
offset = 0;
|
| 42 |
+
Module.setValue(ptr, buffer + offset, 'i8*');
|
| 43 |
offset += modelLen;
|
| 44 |
|
| 45 |
+
Module.setValue(ptr + 4, buffer + offset, 'i8*');
|
| 46 |
offset += lexiconLen;
|
| 47 |
|
| 48 |
+
Module.setValue(ptr + 8, buffer + offset, 'i8*');
|
| 49 |
offset += tokensLen;
|
| 50 |
|
| 51 |
+
Module.setValue(ptr + 12, buffer + offset, 'i8*');
|
| 52 |
offset += dataDirLen;
|
| 53 |
|
| 54 |
+
Module.setValue(ptr + 16, config.noiseScale, 'float');
|
| 55 |
+
Module.setValue(ptr + 20, config.noiseScaleW, 'float');
|
| 56 |
+
Module.setValue(ptr + 24, config.lengthScale, 'float');
|
| 57 |
|
| 58 |
return {
|
| 59 |
buffer: buffer, ptr: ptr, len: len,
|
| 60 |
}
|
| 61 |
}
|
| 62 |
|
| 63 |
+
function initSherpaOnnxOfflineTtsModelConfig(config, Module) {
|
| 64 |
+
const vitsModelConfig = initSherpaOnnxOfflineTtsVitsModelConfig(
|
| 65 |
+
config.offlineTtsVitsModelConfig, Module);
|
| 66 |
|
| 67 |
+
const len = vitsModelConfig.len + 3 * 4;
|
| 68 |
+
const ptr = Module._malloc(len);
|
| 69 |
|
| 70 |
let offset = 0;
|
| 71 |
+
Module._CopyHeap(vitsModelConfig.ptr, vitsModelConfig.len, ptr + offset);
|
| 72 |
offset += vitsModelConfig.len;
|
| 73 |
|
| 74 |
+
Module.setValue(ptr + offset, config.numThreads, 'i32');
|
| 75 |
offset += 4;
|
| 76 |
|
| 77 |
+
Module.setValue(ptr + offset, config.debug, 'i32');
|
| 78 |
offset += 4;
|
| 79 |
|
| 80 |
+
const providerLen = Module.lengthBytesUTF8(config.provider) + 1;
|
| 81 |
+
const buffer = Module._malloc(providerLen);
|
| 82 |
+
Module.stringToUTF8(config.provider, buffer, providerLen);
|
| 83 |
+
Module.setValue(ptr + offset, buffer, 'i8*');
|
| 84 |
|
| 85 |
return {
|
| 86 |
buffer: buffer, ptr: ptr, len: len, config: vitsModelConfig,
|
| 87 |
}
|
| 88 |
}
|
| 89 |
|
| 90 |
+
function initSherpaOnnxOfflineTtsConfig(config, Module) {
|
| 91 |
+
const modelConfig =
|
| 92 |
+
initSherpaOnnxOfflineTtsModelConfig(config.offlineTtsModelConfig, Module);
|
| 93 |
+
const len = modelConfig.len + 2 * 4;
|
| 94 |
+
const ptr = Module._malloc(len);
|
| 95 |
|
| 96 |
let offset = 0;
|
| 97 |
+
Module._CopyHeap(modelConfig.ptr, modelConfig.len, ptr + offset);
|
| 98 |
offset += modelConfig.len;
|
| 99 |
|
| 100 |
+
const ruleFstsLen = Module.lengthBytesUTF8(config.ruleFsts) + 1;
|
| 101 |
+
const buffer = Module._malloc(ruleFstsLen);
|
| 102 |
+
Module.stringToUTF8(config.ruleFsts, buffer, ruleFstsLen);
|
| 103 |
+
Module.setValue(ptr + offset, buffer, 'i8*');
|
| 104 |
offset += 4;
|
| 105 |
|
| 106 |
+
Module.setValue(ptr + offset, config.maxNumSentences, 'i32');
|
| 107 |
|
| 108 |
return {
|
| 109 |
buffer: buffer, ptr: ptr, len: len, config: modelConfig,
|
|
|
|
| 111 |
}
|
| 112 |
|
| 113 |
class OfflineTts {
|
| 114 |
+
constructor(configObj, Module) {
|
| 115 |
+
console.log(configObj)
|
| 116 |
+
const config = initSherpaOnnxOfflineTtsConfig(configObj, Module)
|
| 117 |
+
const handle = Module._SherpaOnnxCreateOfflineTts(config.ptr);
|
| 118 |
|
| 119 |
+
freeConfig(config, Module);
|
| 120 |
|
| 121 |
this.handle = handle;
|
| 122 |
+
this.sampleRate = Module._SherpaOnnxOfflineTtsSampleRate(this.handle);
|
| 123 |
+
this.numSpeakers = Module._SherpaOnnxOfflineTtsNumSpeakers(this.handle);
|
| 124 |
+
this.Module = Module
|
| 125 |
}
|
| 126 |
|
| 127 |
free() {
|
| 128 |
+
this.Module._SherpaOnnxDestroyOfflineTts(this.handle);
|
| 129 |
this.handle = 0
|
| 130 |
}
|
| 131 |
|
|
|
|
| 135 |
// speed: 1.0
|
| 136 |
// }
|
| 137 |
generate(config) {
|
| 138 |
+
const textLen = this.Module.lengthBytesUTF8(config.text) + 1;
|
| 139 |
+
const textPtr = this.Module._malloc(textLen);
|
| 140 |
+
this.Module.stringToUTF8(config.text, textPtr, textLen);
|
| 141 |
|
| 142 |
+
const h = this.Module._SherpaOnnxOfflineTtsGenerate(
|
| 143 |
this.handle, textPtr, config.sid, config.speed);
|
| 144 |
|
| 145 |
+
const numSamples = this.Module.HEAP32[h / 4 + 1];
|
| 146 |
+
const sampleRate = this.Module.HEAP32[h / 4 + 2];
|
| 147 |
|
| 148 |
+
const samplesPtr = this.Module.HEAP32[h / 4] / 4;
|
| 149 |
+
const samples = new Float32Array(numSamples);
|
| 150 |
for (let i = 0; i < numSamples; i++) {
|
| 151 |
+
samples[i] = this.Module.HEAPF32[samplesPtr + i];
|
| 152 |
}
|
| 153 |
|
| 154 |
+
this.Module._SherpaOnnxDestroyOfflineTtsGeneratedAudio(h);
|
| 155 |
return {samples: samples, sampleRate: sampleRate};
|
| 156 |
}
|
| 157 |
+
save(filename, audio) {
|
| 158 |
+
const samples = audio.samples;
|
| 159 |
+
const sampleRate = audio.sampleRate;
|
| 160 |
+
const ptr = this.Module._malloc(samples.length * 4);
|
| 161 |
+
for (let i = 0; i < samples.length; i++) {
|
| 162 |
+
this.Module.HEAPF32[ptr / 4 + i] = samples[i];
|
| 163 |
+
}
|
| 164 |
+
|
| 165 |
+
const filenameLen = this.Module.lengthBytesUTF8(filename) + 1;
|
| 166 |
+
const buffer = this.Module._malloc(filenameLen);
|
| 167 |
+
this.Module.stringToUTF8(filename, buffer, filenameLen);
|
| 168 |
+
this.Module._SherpaOnnxWriteWave(ptr, samples.length, sampleRate, buffer);
|
| 169 |
+
this.Module._free(buffer);
|
| 170 |
+
this.Module._free(ptr);
|
| 171 |
+
}
|
| 172 |
}
|
| 173 |
|
| 174 |
+
function createOfflineTts(Module, myConfig) {
|
| 175 |
+
const offlineTtsVitsModelConfig = {
|
| 176 |
model: './model.onnx',
|
| 177 |
lexicon: '',
|
| 178 |
tokens: './tokens.txt',
|
|
|
|
| 181 |
noiseScaleW: 0.8,
|
| 182 |
lengthScale: 1.0,
|
| 183 |
};
|
| 184 |
+
const offlineTtsModelConfig = {
|
| 185 |
offlineTtsVitsModelConfig: offlineTtsVitsModelConfig,
|
| 186 |
numThreads: 1,
|
| 187 |
debug: 1,
|
|
|
|
| 193 |
maxNumSentences: 1,
|
| 194 |
}
|
| 195 |
|
| 196 |
+
if (myConfig) {
|
| 197 |
+
offlineTtsConfig = myConfig;
|
| 198 |
+
}
|
| 199 |
+
|
| 200 |
+
return new OfflineTts(offlineTtsConfig, Module);
|
| 201 |
+
}
|
| 202 |
+
|
| 203 |
+
if (typeof process == 'object' && typeof process.versions == 'object' &&
|
| 204 |
+
typeof process.versions.node == 'string') {
|
| 205 |
+
module.exports = {
|
| 206 |
+
createOfflineTts,
|
| 207 |
+
};
|
| 208 |
}
|
sherpa-onnx-wasm-main-tts.js
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|