@lee-zg/melange 1.0.0 → 1.2.2
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 +14 -14
- package/dist/chunk-3RM45M64.js +1637 -0
- package/dist/chunk-3RM45M64.js.map +1 -0
- package/dist/{chunk-Q7XG6YN6.cjs → chunk-BEY4UAYF.cjs} +3 -5
- package/dist/chunk-BEY4UAYF.cjs.map +1 -0
- package/dist/{chunk-ALBD5XC5.js → chunk-GXFWPL5M.js} +4 -8
- package/dist/chunk-GXFWPL5M.js.map +1 -0
- package/dist/{chunk-352XNR3C.js → chunk-GZBY4BUP.js} +7 -24
- package/dist/chunk-GZBY4BUP.js.map +1 -0
- package/dist/{chunk-2PXWQDZC.js → chunk-ILNGTDQ4.js} +3 -5
- package/dist/chunk-ILNGTDQ4.js.map +1 -0
- package/dist/{chunk-Q73NOVWX.cjs → chunk-JLBTZPBY.cjs} +7 -24
- package/dist/chunk-JLBTZPBY.cjs.map +1 -0
- package/dist/{chunk-ZT6HVG4G.cjs → chunk-UYJUSNDI.cjs} +4 -8
- package/dist/chunk-UYJUSNDI.cjs.map +1 -0
- package/dist/chunk-YZVCK6VZ.cjs +1646 -0
- package/dist/chunk-YZVCK6VZ.cjs.map +1 -0
- package/dist/core/index.cjs +23 -23
- package/dist/core/index.js +1 -1
- package/dist/fp/index.cjs +45 -45
- package/dist/fp/index.js +1 -1
- package/dist/index.cjs +147 -147
- package/dist/index.js +4 -4
- package/dist/plugins/index.cjs +9 -9
- package/dist/plugins/index.d.cts +287 -133
- package/dist/plugins/index.d.ts +287 -133
- package/dist/plugins/index.js +1 -1
- package/dist/utils/index.cjs +73 -73
- package/dist/utils/index.js +1 -1
- package/package.json +5 -2
- package/dist/chunk-2PXWQDZC.js.map +0 -1
- package/dist/chunk-352XNR3C.js.map +0 -1
- package/dist/chunk-ALBD5XC5.js.map +0 -1
- package/dist/chunk-O7K662J5.cjs +0 -842
- package/dist/chunk-O7K662J5.cjs.map +0 -1
- package/dist/chunk-Q73NOVWX.cjs.map +0 -1
- package/dist/chunk-Q7XG6YN6.cjs.map +0 -1
- package/dist/chunk-YGMBCZJQ.js +0 -833
- package/dist/chunk-YGMBCZJQ.js.map +0 -1
- package/dist/chunk-ZT6HVG4G.cjs.map +0 -1
|
@@ -0,0 +1,1646 @@
|
|
|
1
|
+
'use strict';
|
|
2
|
+
|
|
3
|
+
var chunkPK6SKIKE_cjs = require('./chunk-PK6SKIKE.cjs');
|
|
4
|
+
|
|
5
|
+
// src/plugins/speech/synthesis.ts
|
|
6
|
+
var SynthesisAudioUtils = {
|
|
7
|
+
/**
|
|
8
|
+
* 创建 AudioContext
|
|
9
|
+
*/
|
|
10
|
+
createAudioContext() {
|
|
11
|
+
const AudioCtor = typeof window !== "undefined" ? window.AudioContext || window.webkitAudioContext : null;
|
|
12
|
+
if (!AudioCtor) {
|
|
13
|
+
throw new Error("\u6D4F\u89C8\u5668\u4E0D\u652F\u6301 AudioContext");
|
|
14
|
+
}
|
|
15
|
+
return new AudioCtor();
|
|
16
|
+
},
|
|
17
|
+
/**
|
|
18
|
+
* ArrayBuffer 转 Base64
|
|
19
|
+
*/
|
|
20
|
+
arrayBufferToBase64(buffer) {
|
|
21
|
+
let binary = "";
|
|
22
|
+
const bytes = new Uint8Array(buffer);
|
|
23
|
+
for (let i = 0; i < bytes.byteLength; i++) {
|
|
24
|
+
binary += String.fromCharCode(bytes[i] ?? 0);
|
|
25
|
+
}
|
|
26
|
+
return typeof btoa !== "undefined" ? btoa(binary) : Buffer.from(binary).toString("base64");
|
|
27
|
+
},
|
|
28
|
+
/**
|
|
29
|
+
* Base64 转 ArrayBuffer
|
|
30
|
+
*/
|
|
31
|
+
base64ToArrayBuffer(base64) {
|
|
32
|
+
const binary = typeof atob !== "undefined" ? atob(base64) : Buffer.from(base64, "base64").toString("binary");
|
|
33
|
+
const bytes = new Uint8Array(binary.length);
|
|
34
|
+
for (let i = 0; i < binary.length; i++) {
|
|
35
|
+
bytes[i] = binary.charCodeAt(i);
|
|
36
|
+
}
|
|
37
|
+
return bytes.buffer;
|
|
38
|
+
},
|
|
39
|
+
/**
|
|
40
|
+
* 计算音频时长(粗略估算)
|
|
41
|
+
*/
|
|
42
|
+
estimateDuration(byteLength, format) {
|
|
43
|
+
const bitRates = {
|
|
44
|
+
mp3: 128e3,
|
|
45
|
+
// 128 kbps
|
|
46
|
+
wav: 256e3,
|
|
47
|
+
// 16bit 16kHz mono
|
|
48
|
+
ogg: 96e3,
|
|
49
|
+
// 96 kbps
|
|
50
|
+
pcm: 256e3
|
|
51
|
+
// 16bit 16kHz mono
|
|
52
|
+
};
|
|
53
|
+
return Math.ceil(byteLength * 8 / bitRates[format] * 1e3);
|
|
54
|
+
}
|
|
55
|
+
};
|
|
56
|
+
var BaseSynthesisStrategy = class {
|
|
57
|
+
constructor(config) {
|
|
58
|
+
this.config = config;
|
|
59
|
+
}
|
|
60
|
+
static {
|
|
61
|
+
chunkPK6SKIKE_cjs.__name(this, "BaseSynthesisStrategy");
|
|
62
|
+
}
|
|
63
|
+
listeners = {
|
|
64
|
+
start: [],
|
|
65
|
+
end: [],
|
|
66
|
+
pause: [],
|
|
67
|
+
resume: [],
|
|
68
|
+
boundary: [],
|
|
69
|
+
mark: [],
|
|
70
|
+
error: [],
|
|
71
|
+
state: []
|
|
72
|
+
};
|
|
73
|
+
_status = "IDLE" /* IDLE */;
|
|
74
|
+
/**
|
|
75
|
+
* 获取当前状态
|
|
76
|
+
*/
|
|
77
|
+
get status() {
|
|
78
|
+
return this._status;
|
|
79
|
+
}
|
|
80
|
+
/**
|
|
81
|
+
* 设置状态
|
|
82
|
+
*/
|
|
83
|
+
setStatus(status) {
|
|
84
|
+
if (this._status !== status) {
|
|
85
|
+
this._status = status;
|
|
86
|
+
this.emit("state", status);
|
|
87
|
+
}
|
|
88
|
+
}
|
|
89
|
+
/**
|
|
90
|
+
* 添加事件监听
|
|
91
|
+
*/
|
|
92
|
+
on(event, fn) {
|
|
93
|
+
this.listeners[event].push(fn);
|
|
94
|
+
}
|
|
95
|
+
/**
|
|
96
|
+
* 移除事件监听
|
|
97
|
+
*/
|
|
98
|
+
off(event, fn) {
|
|
99
|
+
const listeners = this.listeners[event];
|
|
100
|
+
const index = listeners.indexOf(fn);
|
|
101
|
+
if (index > -1) {
|
|
102
|
+
listeners.splice(index, 1);
|
|
103
|
+
}
|
|
104
|
+
}
|
|
105
|
+
/**
|
|
106
|
+
* 触发事件
|
|
107
|
+
*/
|
|
108
|
+
emit(event, data) {
|
|
109
|
+
const listeners = this.listeners[event];
|
|
110
|
+
listeners.forEach((fn) => {
|
|
111
|
+
try {
|
|
112
|
+
fn(data);
|
|
113
|
+
} catch (e) {
|
|
114
|
+
console.error(`[Synthesis] \u4E8B\u4EF6\u5904\u7406\u5668\u9519\u8BEF (${event}):`, e);
|
|
115
|
+
}
|
|
116
|
+
});
|
|
117
|
+
}
|
|
118
|
+
/**
|
|
119
|
+
* 销毁资源
|
|
120
|
+
*/
|
|
121
|
+
dispose() {
|
|
122
|
+
this.cancel();
|
|
123
|
+
Object.keys(this.listeners).forEach((key) => {
|
|
124
|
+
this.listeners[key] = [];
|
|
125
|
+
});
|
|
126
|
+
}
|
|
127
|
+
};
|
|
128
|
+
var NativeSynthesisStrategy = class extends BaseSynthesisStrategy {
|
|
129
|
+
static {
|
|
130
|
+
chunkPK6SKIKE_cjs.__name(this, "NativeSynthesisStrategy");
|
|
131
|
+
}
|
|
132
|
+
synthesis = null;
|
|
133
|
+
// Internal state tracking for current utterance (write-only, used for GC reference)
|
|
134
|
+
currentUtterance = null;
|
|
135
|
+
voicesLoaded = false;
|
|
136
|
+
voicesPromise = null;
|
|
137
|
+
constructor(config) {
|
|
138
|
+
super(config);
|
|
139
|
+
if (this.isAvailable()) {
|
|
140
|
+
this.synthesis = window.speechSynthesis;
|
|
141
|
+
}
|
|
142
|
+
}
|
|
143
|
+
isAvailable() {
|
|
144
|
+
return typeof window !== "undefined" && "speechSynthesis" in window;
|
|
145
|
+
}
|
|
146
|
+
async getVoices() {
|
|
147
|
+
if (!this.synthesis) return [];
|
|
148
|
+
if (this.voicesPromise) {
|
|
149
|
+
return this.voicesPromise;
|
|
150
|
+
}
|
|
151
|
+
this.voicesPromise = new Promise((resolve) => {
|
|
152
|
+
const loadVoices = /* @__PURE__ */ chunkPK6SKIKE_cjs.__name(() => {
|
|
153
|
+
const voices = this.synthesis.getVoices();
|
|
154
|
+
if (voices.length > 0) {
|
|
155
|
+
this.voicesLoaded = true;
|
|
156
|
+
resolve(
|
|
157
|
+
voices.map((voice) => ({
|
|
158
|
+
id: voice.voiceURI,
|
|
159
|
+
name: voice.name,
|
|
160
|
+
lang: voice.lang,
|
|
161
|
+
localService: voice.localService,
|
|
162
|
+
default: voice.default,
|
|
163
|
+
provider: "browser"
|
|
164
|
+
}))
|
|
165
|
+
);
|
|
166
|
+
}
|
|
167
|
+
}, "loadVoices");
|
|
168
|
+
loadVoices();
|
|
169
|
+
if (!this.voicesLoaded) {
|
|
170
|
+
this.synthesis.addEventListener("voiceschanged", loadVoices, { once: true });
|
|
171
|
+
setTimeout(() => {
|
|
172
|
+
if (!this.voicesLoaded) {
|
|
173
|
+
resolve([]);
|
|
174
|
+
}
|
|
175
|
+
}, 3e3);
|
|
176
|
+
}
|
|
177
|
+
});
|
|
178
|
+
return this.voicesPromise;
|
|
179
|
+
}
|
|
180
|
+
async speak(text, config) {
|
|
181
|
+
if (!this.synthesis) {
|
|
182
|
+
throw { code: "NOT_SUPPORTED", message: "\u6D4F\u89C8\u5668\u4E0D\u652F\u6301\u8BED\u97F3\u5408\u6210" };
|
|
183
|
+
}
|
|
184
|
+
this.cancel();
|
|
185
|
+
this.setStatus("LOADING" /* LOADING */);
|
|
186
|
+
return new Promise((resolve, reject) => {
|
|
187
|
+
const mergedConfig = { ...this.config, ...config };
|
|
188
|
+
const utterance = new SpeechSynthesisUtterance(text);
|
|
189
|
+
this.currentUtterance = utterance;
|
|
190
|
+
if (mergedConfig.lang) utterance.lang = mergedConfig.lang;
|
|
191
|
+
if (mergedConfig.volume !== void 0) utterance.volume = mergedConfig.volume;
|
|
192
|
+
if (mergedConfig.rate !== void 0) utterance.rate = mergedConfig.rate;
|
|
193
|
+
if (mergedConfig.pitch !== void 0) utterance.pitch = mergedConfig.pitch;
|
|
194
|
+
if (mergedConfig.voice) {
|
|
195
|
+
const voices = this.synthesis.getVoices();
|
|
196
|
+
const voiceConfig = mergedConfig.voice;
|
|
197
|
+
const targetVoice = typeof voiceConfig === "string" ? voices.find((v) => v.name === voiceConfig || v.voiceURI === voiceConfig) : voices.find((v) => v.voiceURI === voiceConfig.id);
|
|
198
|
+
if (targetVoice) {
|
|
199
|
+
utterance.voice = targetVoice;
|
|
200
|
+
}
|
|
201
|
+
}
|
|
202
|
+
utterance.onstart = () => {
|
|
203
|
+
this.setStatus("SPEAKING" /* SPEAKING */);
|
|
204
|
+
this.emit("start");
|
|
205
|
+
};
|
|
206
|
+
utterance.onend = () => {
|
|
207
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
208
|
+
this.currentUtterance = null;
|
|
209
|
+
this.emit("end");
|
|
210
|
+
resolve();
|
|
211
|
+
};
|
|
212
|
+
utterance.onpause = () => {
|
|
213
|
+
this.setStatus("PAUSED" /* PAUSED */);
|
|
214
|
+
this.emit("pause");
|
|
215
|
+
};
|
|
216
|
+
utterance.onresume = () => {
|
|
217
|
+
this.setStatus("SPEAKING" /* SPEAKING */);
|
|
218
|
+
this.emit("resume");
|
|
219
|
+
};
|
|
220
|
+
utterance.onboundary = (event) => {
|
|
221
|
+
this.emit("boundary", {
|
|
222
|
+
charIndex: event.charIndex,
|
|
223
|
+
charLength: event.charLength,
|
|
224
|
+
elapsedTime: event.elapsedTime,
|
|
225
|
+
name: event.name
|
|
226
|
+
});
|
|
227
|
+
};
|
|
228
|
+
utterance.onmark = (event) => {
|
|
229
|
+
this.emit("mark", { name: event.name });
|
|
230
|
+
};
|
|
231
|
+
utterance.onerror = (event) => {
|
|
232
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
233
|
+
this.currentUtterance = null;
|
|
234
|
+
const error = {
|
|
235
|
+
code: "UNKNOWN",
|
|
236
|
+
message: `\u8BED\u97F3\u5408\u6210\u9519\u8BEF: ${event.error}`,
|
|
237
|
+
originalError: event
|
|
238
|
+
};
|
|
239
|
+
this.emit("error", error);
|
|
240
|
+
reject(error);
|
|
241
|
+
};
|
|
242
|
+
this.synthesis.speak(utterance);
|
|
243
|
+
});
|
|
244
|
+
}
|
|
245
|
+
pause() {
|
|
246
|
+
if (this.synthesis && this.isSpeaking()) {
|
|
247
|
+
this.synthesis.pause();
|
|
248
|
+
}
|
|
249
|
+
}
|
|
250
|
+
resume() {
|
|
251
|
+
if (this.synthesis && this.isPaused()) {
|
|
252
|
+
this.synthesis.resume();
|
|
253
|
+
}
|
|
254
|
+
}
|
|
255
|
+
cancel() {
|
|
256
|
+
if (this.synthesis) {
|
|
257
|
+
this.synthesis.cancel();
|
|
258
|
+
this.currentUtterance = null;
|
|
259
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
260
|
+
}
|
|
261
|
+
}
|
|
262
|
+
isSpeaking() {
|
|
263
|
+
return this.currentUtterance !== null && (this.synthesis?.speaking ?? false);
|
|
264
|
+
}
|
|
265
|
+
isPaused() {
|
|
266
|
+
return this.synthesis?.paused ?? false;
|
|
267
|
+
}
|
|
268
|
+
};
|
|
269
|
+
var CloudSynthesisStrategy = class extends BaseSynthesisStrategy {
|
|
270
|
+
static {
|
|
271
|
+
chunkPK6SKIKE_cjs.__name(this, "CloudSynthesisStrategy");
|
|
272
|
+
}
|
|
273
|
+
adapter;
|
|
274
|
+
audioContext = null;
|
|
275
|
+
sourceNode = null;
|
|
276
|
+
gainNode = null;
|
|
277
|
+
startTime = 0;
|
|
278
|
+
pauseTime = 0;
|
|
279
|
+
audioBuffer = null;
|
|
280
|
+
constructor(config) {
|
|
281
|
+
super(config);
|
|
282
|
+
if (!config.cloudAdapter) {
|
|
283
|
+
throw new Error("[CloudStrategy] \u9700\u8981\u63D0\u4F9B cloudAdapter");
|
|
284
|
+
}
|
|
285
|
+
this.adapter = config.cloudAdapter;
|
|
286
|
+
}
|
|
287
|
+
isAvailable() {
|
|
288
|
+
return typeof window !== "undefined" && "AudioContext" in window;
|
|
289
|
+
}
|
|
290
|
+
async getVoices() {
|
|
291
|
+
if (!this.adapter.getVoices) return [];
|
|
292
|
+
try {
|
|
293
|
+
const cloudVoices = await this.adapter.getVoices();
|
|
294
|
+
return cloudVoices.map((v) => ({
|
|
295
|
+
id: v.id,
|
|
296
|
+
name: v.name,
|
|
297
|
+
lang: v.lang,
|
|
298
|
+
localService: false,
|
|
299
|
+
default: false,
|
|
300
|
+
provider: "custom"
|
|
301
|
+
}));
|
|
302
|
+
} catch {
|
|
303
|
+
return [];
|
|
304
|
+
}
|
|
305
|
+
}
|
|
306
|
+
async speak(text, config) {
|
|
307
|
+
if (!text.trim()) {
|
|
308
|
+
throw { code: "INVALID_TEXT", message: "\u6587\u672C\u5185\u5BB9\u4E3A\u7A7A" };
|
|
309
|
+
}
|
|
310
|
+
this.cancel();
|
|
311
|
+
const mergedConfig = { ...this.config, ...config };
|
|
312
|
+
this.setStatus("LOADING" /* LOADING */);
|
|
313
|
+
try {
|
|
314
|
+
this.emit("start");
|
|
315
|
+
const result = await this.adapter.synthesize(text, mergedConfig);
|
|
316
|
+
if (!this.audioContext) {
|
|
317
|
+
this.audioContext = SynthesisAudioUtils.createAudioContext();
|
|
318
|
+
}
|
|
319
|
+
if (this.audioContext.state === "suspended") {
|
|
320
|
+
await this.audioContext.resume();
|
|
321
|
+
}
|
|
322
|
+
this.audioBuffer = await this.audioContext.decodeAudioData(result.audioData.slice(0));
|
|
323
|
+
await this.playBuffer();
|
|
324
|
+
} catch (e) {
|
|
325
|
+
const error = e;
|
|
326
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
327
|
+
const synthError = {
|
|
328
|
+
code: "ADAPTER_ERROR",
|
|
329
|
+
message: error.message || "\u5408\u6210\u5931\u8D25",
|
|
330
|
+
originalError: e
|
|
331
|
+
};
|
|
332
|
+
this.emit("error", synthError);
|
|
333
|
+
throw synthError;
|
|
334
|
+
}
|
|
335
|
+
}
|
|
336
|
+
async playBuffer() {
|
|
337
|
+
if (!this.audioContext || !this.audioBuffer) return;
|
|
338
|
+
return new Promise((resolve, reject) => {
|
|
339
|
+
try {
|
|
340
|
+
this.sourceNode = this.audioContext.createBufferSource();
|
|
341
|
+
this.sourceNode.buffer = this.audioBuffer;
|
|
342
|
+
this.gainNode = this.audioContext.createGain();
|
|
343
|
+
this.gainNode.gain.value = this.config.volume ?? 1;
|
|
344
|
+
this.sourceNode.connect(this.gainNode);
|
|
345
|
+
this.gainNode.connect(this.audioContext.destination);
|
|
346
|
+
this.sourceNode.onended = () => {
|
|
347
|
+
if (this._status === "SPEAKING" /* SPEAKING */) {
|
|
348
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
349
|
+
this.emit("end");
|
|
350
|
+
resolve();
|
|
351
|
+
}
|
|
352
|
+
};
|
|
353
|
+
const offset = this.pauseTime > 0 ? this.pauseTime : 0;
|
|
354
|
+
this.sourceNode.start(0, offset);
|
|
355
|
+
this.startTime = this.audioContext.currentTime - offset;
|
|
356
|
+
this.pauseTime = 0;
|
|
357
|
+
this.setStatus("SPEAKING" /* SPEAKING */);
|
|
358
|
+
} catch (e) {
|
|
359
|
+
reject(e);
|
|
360
|
+
}
|
|
361
|
+
});
|
|
362
|
+
}
|
|
363
|
+
pause() {
|
|
364
|
+
if (this._status === "SPEAKING" /* SPEAKING */ && this.audioContext && this.sourceNode) {
|
|
365
|
+
this.pauseTime = this.audioContext.currentTime - this.startTime;
|
|
366
|
+
this.sourceNode.stop();
|
|
367
|
+
this.sourceNode.disconnect();
|
|
368
|
+
this.sourceNode = null;
|
|
369
|
+
this.setStatus("PAUSED" /* PAUSED */);
|
|
370
|
+
this.emit("pause");
|
|
371
|
+
}
|
|
372
|
+
}
|
|
373
|
+
resume() {
|
|
374
|
+
if (this._status === "PAUSED" /* PAUSED */ && this.audioBuffer) {
|
|
375
|
+
this.emit("resume");
|
|
376
|
+
void this.playBuffer();
|
|
377
|
+
}
|
|
378
|
+
}
|
|
379
|
+
cancel() {
|
|
380
|
+
if (this.sourceNode) {
|
|
381
|
+
try {
|
|
382
|
+
this.sourceNode.stop();
|
|
383
|
+
this.sourceNode.disconnect();
|
|
384
|
+
} catch {
|
|
385
|
+
}
|
|
386
|
+
this.sourceNode = null;
|
|
387
|
+
}
|
|
388
|
+
this.audioBuffer = null;
|
|
389
|
+
this.pauseTime = 0;
|
|
390
|
+
this.startTime = 0;
|
|
391
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
392
|
+
}
|
|
393
|
+
isSpeaking() {
|
|
394
|
+
return this._status === "SPEAKING" /* SPEAKING */;
|
|
395
|
+
}
|
|
396
|
+
isPaused() {
|
|
397
|
+
return this._status === "PAUSED" /* PAUSED */;
|
|
398
|
+
}
|
|
399
|
+
dispose() {
|
|
400
|
+
super.dispose();
|
|
401
|
+
this.cancel();
|
|
402
|
+
if (this.audioContext) {
|
|
403
|
+
void this.audioContext.close();
|
|
404
|
+
this.audioContext = null;
|
|
405
|
+
}
|
|
406
|
+
}
|
|
407
|
+
};
|
|
408
|
+
var SpeechSynthesizerImpl = class {
|
|
409
|
+
static {
|
|
410
|
+
chunkPK6SKIKE_cjs.__name(this, "SpeechSynthesizerImpl");
|
|
411
|
+
}
|
|
412
|
+
strategy = null;
|
|
413
|
+
config = {};
|
|
414
|
+
_currentProvider = "browser";
|
|
415
|
+
_status = "idle";
|
|
416
|
+
eventHandlers = /* @__PURE__ */ new Map();
|
|
417
|
+
customProviders = /* @__PURE__ */ new Map();
|
|
418
|
+
get currentProvider() {
|
|
419
|
+
return this._currentProvider;
|
|
420
|
+
}
|
|
421
|
+
get status() {
|
|
422
|
+
return this._status;
|
|
423
|
+
}
|
|
424
|
+
/**
|
|
425
|
+
* 获取当前合成状态
|
|
426
|
+
*/
|
|
427
|
+
get synthesisStatus() {
|
|
428
|
+
return this.strategy?.status ?? "IDLE" /* IDLE */;
|
|
429
|
+
}
|
|
430
|
+
/**
|
|
431
|
+
* 初始化语音合成器
|
|
432
|
+
*/
|
|
433
|
+
initialize(config) {
|
|
434
|
+
this._status = "loading";
|
|
435
|
+
this.config = {
|
|
436
|
+
lang: "zh-CN",
|
|
437
|
+
volume: 1,
|
|
438
|
+
rate: 1,
|
|
439
|
+
pitch: 1,
|
|
440
|
+
preferredProvider: "browser",
|
|
441
|
+
autoFallback: true,
|
|
442
|
+
fallbackProviders: ["azure", "google", "aws"],
|
|
443
|
+
mode: "auto",
|
|
444
|
+
audioFormat: "mp3",
|
|
445
|
+
...config
|
|
446
|
+
};
|
|
447
|
+
try {
|
|
448
|
+
this.initializeStrategy();
|
|
449
|
+
this._status = "ready";
|
|
450
|
+
return Promise.resolve();
|
|
451
|
+
} catch (error) {
|
|
452
|
+
this._status = "error";
|
|
453
|
+
return Promise.reject(error);
|
|
454
|
+
}
|
|
455
|
+
}
|
|
456
|
+
/**
|
|
457
|
+
* 初始化合成策略
|
|
458
|
+
*/
|
|
459
|
+
initializeStrategy() {
|
|
460
|
+
const advConfig = this.config;
|
|
461
|
+
const mode = advConfig.mode ?? "auto";
|
|
462
|
+
const hasNative = typeof window !== "undefined" && "speechSynthesis" in window;
|
|
463
|
+
if (mode === "native" || mode === "auto" && hasNative && !advConfig.cloudAdapter) {
|
|
464
|
+
const nativeStrategy = new NativeSynthesisStrategy(advConfig);
|
|
465
|
+
if (!nativeStrategy.isAvailable()) {
|
|
466
|
+
throw new Error("\u6D4F\u89C8\u5668\u4E0D\u652F\u6301\u539F\u751F\u8BED\u97F3\u5408\u6210");
|
|
467
|
+
}
|
|
468
|
+
this.strategy = nativeStrategy;
|
|
469
|
+
this._currentProvider = "browser";
|
|
470
|
+
} else if (mode === "cloud" || mode === "auto" && advConfig.cloudAdapter) {
|
|
471
|
+
if (!advConfig.cloudAdapter) {
|
|
472
|
+
throw new Error("\u4E91\u7AEF\u6A21\u5F0F\u9700\u8981\u63D0\u4F9B cloudAdapter");
|
|
473
|
+
}
|
|
474
|
+
this.strategy = new CloudSynthesisStrategy(advConfig);
|
|
475
|
+
this._currentProvider = "custom";
|
|
476
|
+
} else {
|
|
477
|
+
throw new Error("\u6CA1\u6709\u53EF\u7528\u7684\u5408\u6210\u7B56\u7565");
|
|
478
|
+
}
|
|
479
|
+
this.forwardStrategyEvents();
|
|
480
|
+
}
|
|
481
|
+
/**
|
|
482
|
+
* 转发策略事件
|
|
483
|
+
*/
|
|
484
|
+
forwardStrategyEvents() {
|
|
485
|
+
if (!this.strategy) return;
|
|
486
|
+
this.strategy.on("start", () => this.emit("start", { type: "start" }));
|
|
487
|
+
this.strategy.on("end", () => this.emit("end", { type: "end" }));
|
|
488
|
+
this.strategy.on("pause", () => this.emit("pause", { type: "pause" }));
|
|
489
|
+
this.strategy.on("resume", () => this.emit("resume", { type: "resume" }));
|
|
490
|
+
this.strategy.on("boundary", (data) => {
|
|
491
|
+
const event = { type: "boundary" };
|
|
492
|
+
if (data.charIndex !== void 0) event.charIndex = data.charIndex;
|
|
493
|
+
if (data.charLength !== void 0) event.charLength = data.charLength;
|
|
494
|
+
if (data.elapsedTime !== void 0) event.elapsedTime = data.elapsedTime;
|
|
495
|
+
if (data.name !== void 0) event.name = data.name;
|
|
496
|
+
this.emit("boundary", event);
|
|
497
|
+
});
|
|
498
|
+
this.strategy.on("mark", (data) => {
|
|
499
|
+
const event = { type: "mark" };
|
|
500
|
+
if (data.name !== void 0) event.name = data.name;
|
|
501
|
+
this.emit("mark", event);
|
|
502
|
+
});
|
|
503
|
+
this.strategy.on("error", (err) => {
|
|
504
|
+
const errorEvent = {
|
|
505
|
+
type: "error",
|
|
506
|
+
error: {
|
|
507
|
+
code: err.code,
|
|
508
|
+
message: err.message
|
|
509
|
+
}
|
|
510
|
+
};
|
|
511
|
+
if (err.originalError) {
|
|
512
|
+
errorEvent.error.originalError = err.originalError;
|
|
513
|
+
}
|
|
514
|
+
this.emit("error", errorEvent);
|
|
515
|
+
});
|
|
516
|
+
}
|
|
517
|
+
/**
|
|
518
|
+
* 获取可用语音列表
|
|
519
|
+
*/
|
|
520
|
+
async getVoices() {
|
|
521
|
+
if (!this.strategy) {
|
|
522
|
+
throw new Error("\u8BED\u97F3\u5408\u6210\u5668\u672A\u521D\u59CB\u5316");
|
|
523
|
+
}
|
|
524
|
+
return this.strategy.getVoices();
|
|
525
|
+
}
|
|
526
|
+
/**
|
|
527
|
+
* 朗读文本
|
|
528
|
+
*/
|
|
529
|
+
async speak(text, config) {
|
|
530
|
+
if (this._status !== "ready") {
|
|
531
|
+
throw new Error("\u8BED\u97F3\u5408\u6210\u5668\u672A\u5C31\u7EEA");
|
|
532
|
+
}
|
|
533
|
+
return this.strategy?.speak(text, config);
|
|
534
|
+
}
|
|
535
|
+
/**
|
|
536
|
+
* 暂停朗读
|
|
537
|
+
*/
|
|
538
|
+
pause() {
|
|
539
|
+
this.strategy?.pause();
|
|
540
|
+
}
|
|
541
|
+
/**
|
|
542
|
+
* 继续朗读
|
|
543
|
+
*/
|
|
544
|
+
resume() {
|
|
545
|
+
this.strategy?.resume();
|
|
546
|
+
}
|
|
547
|
+
/**
|
|
548
|
+
* 取消朗读
|
|
549
|
+
*/
|
|
550
|
+
cancel() {
|
|
551
|
+
this.strategy?.cancel();
|
|
552
|
+
}
|
|
553
|
+
/**
|
|
554
|
+
* 是否正在朗读
|
|
555
|
+
*/
|
|
556
|
+
isSpeaking() {
|
|
557
|
+
return this.strategy?.isSpeaking() ?? false;
|
|
558
|
+
}
|
|
559
|
+
/**
|
|
560
|
+
* 是否已暂停
|
|
561
|
+
*/
|
|
562
|
+
isPaused() {
|
|
563
|
+
return this.strategy?.isPaused() ?? false;
|
|
564
|
+
}
|
|
565
|
+
/**
|
|
566
|
+
* 添加事件监听
|
|
567
|
+
*/
|
|
568
|
+
on(event, handler) {
|
|
569
|
+
if (!this.eventHandlers.has(event)) {
|
|
570
|
+
this.eventHandlers.set(event, /* @__PURE__ */ new Set());
|
|
571
|
+
}
|
|
572
|
+
this.eventHandlers.get(event).add(handler);
|
|
573
|
+
}
|
|
574
|
+
/**
|
|
575
|
+
* 移除事件监听
|
|
576
|
+
*/
|
|
577
|
+
off(event, handler) {
|
|
578
|
+
this.eventHandlers.get(event)?.delete(handler);
|
|
579
|
+
}
|
|
580
|
+
/**
|
|
581
|
+
* 触发事件
|
|
582
|
+
*/
|
|
583
|
+
emit(type, event) {
|
|
584
|
+
this.eventHandlers.get(type)?.forEach((handler) => {
|
|
585
|
+
try {
|
|
586
|
+
handler(event);
|
|
587
|
+
} catch (e) {
|
|
588
|
+
console.error("[SpeechSynthesizer] \u4E8B\u4EF6\u5904\u7406\u5668\u9519\u8BEF:", e);
|
|
589
|
+
}
|
|
590
|
+
});
|
|
591
|
+
}
|
|
592
|
+
/**
|
|
593
|
+
* 销毁实例
|
|
594
|
+
*/
|
|
595
|
+
dispose() {
|
|
596
|
+
this.strategy?.dispose();
|
|
597
|
+
this.strategy = null;
|
|
598
|
+
this.eventHandlers.clear();
|
|
599
|
+
this._status = "idle";
|
|
600
|
+
}
|
|
601
|
+
/**
|
|
602
|
+
* 注册自定义提供商
|
|
603
|
+
*/
|
|
604
|
+
registerProvider(type, provider) {
|
|
605
|
+
this.customProviders.set(type, provider);
|
|
606
|
+
}
|
|
607
|
+
/**
|
|
608
|
+
* 使用云端适配器
|
|
609
|
+
*/
|
|
610
|
+
useCloudAdapter(adapter) {
|
|
611
|
+
this.config.cloudAdapter = adapter;
|
|
612
|
+
this.config.mode = "cloud";
|
|
613
|
+
}
|
|
614
|
+
};
|
|
615
|
+
async function createSpeechSynthesizer(config) {
|
|
616
|
+
const synthesizer = new SpeechSynthesizerImpl();
|
|
617
|
+
await synthesizer.initialize(config);
|
|
618
|
+
return synthesizer;
|
|
619
|
+
}
|
|
620
|
+
chunkPK6SKIKE_cjs.__name(createSpeechSynthesizer, "createSpeechSynthesizer");
|
|
621
|
+
function isSpeechSynthesisSupported() {
|
|
622
|
+
return typeof window !== "undefined" && "speechSynthesis" in window;
|
|
623
|
+
}
|
|
624
|
+
chunkPK6SKIKE_cjs.__name(isSpeechSynthesisSupported, "isSpeechSynthesisSupported");
|
|
625
|
+
async function speak(text, config) {
|
|
626
|
+
const synthesizer = await createSpeechSynthesizer(config);
|
|
627
|
+
try {
|
|
628
|
+
await synthesizer.speak(text, config);
|
|
629
|
+
} finally {
|
|
630
|
+
synthesizer.dispose();
|
|
631
|
+
}
|
|
632
|
+
}
|
|
633
|
+
chunkPK6SKIKE_cjs.__name(speak, "speak");
|
|
634
|
+
|
|
635
|
+
// src/plugins/speech/recognition.ts
|
|
636
|
+
var AudioUtils = {
|
|
637
|
+
/**
|
|
638
|
+
* 重采样音频数据
|
|
639
|
+
* @param data 原始音频数据
|
|
640
|
+
* @param inputRate 输入采样率
|
|
641
|
+
* @param outputRate 输出采样率
|
|
642
|
+
* @returns 重采样后的数据
|
|
643
|
+
*/
|
|
644
|
+
resample(data, inputRate, outputRate) {
|
|
645
|
+
if (inputRate === outputRate) return data;
|
|
646
|
+
const compression = inputRate / outputRate;
|
|
647
|
+
const length = Math.ceil(data.length / compression);
|
|
648
|
+
const result = new Float32Array(length);
|
|
649
|
+
for (let i = 0; i < length; i++) {
|
|
650
|
+
result[i] = data[Math.floor(i * compression)] ?? 0;
|
|
651
|
+
}
|
|
652
|
+
return result;
|
|
653
|
+
},
|
|
654
|
+
/**
|
|
655
|
+
* Float32 转 Int16 PCM
|
|
656
|
+
* @param input Float32 数据
|
|
657
|
+
* @returns Int16 PCM 数据
|
|
658
|
+
*/
|
|
659
|
+
floatTo16BitPCM(input) {
|
|
660
|
+
const output = new Int16Array(input.length);
|
|
661
|
+
for (let i = 0; i < input.length; i++) {
|
|
662
|
+
const s = Math.max(-1, Math.min(1, input[i] ?? 0));
|
|
663
|
+
output[i] = s < 0 ? s * 32768 : s * 32767;
|
|
664
|
+
}
|
|
665
|
+
return output;
|
|
666
|
+
},
|
|
667
|
+
/**
|
|
668
|
+
* 计算音量 RMS
|
|
669
|
+
* @param data 音频数据
|
|
670
|
+
* @returns RMS 值
|
|
671
|
+
*/
|
|
672
|
+
calculateRMS(data) {
|
|
673
|
+
let sum = 0;
|
|
674
|
+
for (let i = 0; i < data.length; i++) {
|
|
675
|
+
const sample = data[i] ?? 0;
|
|
676
|
+
sum += sample * sample;
|
|
677
|
+
}
|
|
678
|
+
return Math.sqrt(sum / data.length);
|
|
679
|
+
},
|
|
680
|
+
/**
|
|
681
|
+
* 合并 PCM 片段
|
|
682
|
+
* @param buffers PCM 缓冲区数组
|
|
683
|
+
* @param totalLength 总采样数
|
|
684
|
+
* @returns 合并后的 Int16 数组
|
|
685
|
+
*/
|
|
686
|
+
mergeBuffers(buffers, totalLength) {
|
|
687
|
+
const result = new Int16Array(totalLength);
|
|
688
|
+
let offset = 0;
|
|
689
|
+
for (const buffer of buffers) {
|
|
690
|
+
const view = new Int16Array(buffer);
|
|
691
|
+
result.set(view, offset);
|
|
692
|
+
offset += view.length;
|
|
693
|
+
}
|
|
694
|
+
return result;
|
|
695
|
+
},
|
|
696
|
+
/**
|
|
697
|
+
* PCM 转 WAV 封装
|
|
698
|
+
* @param samples PCM 采样数据
|
|
699
|
+
* @param sampleRate 采样率
|
|
700
|
+
* @param channels 声道数
|
|
701
|
+
* @returns WAV 格式 ArrayBuffer
|
|
702
|
+
*/
|
|
703
|
+
encodeWAV(samples, sampleRate = 16e3, channels = 1) {
|
|
704
|
+
const buffer = new ArrayBuffer(44 + samples.length * 2);
|
|
705
|
+
const view = new DataView(buffer);
|
|
706
|
+
const writeString = /* @__PURE__ */ chunkPK6SKIKE_cjs.__name((offset, str) => {
|
|
707
|
+
for (let i = 0; i < str.length; i++) {
|
|
708
|
+
view.setUint8(offset + i, str.charCodeAt(i));
|
|
709
|
+
}
|
|
710
|
+
}, "writeString");
|
|
711
|
+
writeString(0, "RIFF");
|
|
712
|
+
view.setUint32(4, 36 + samples.length * 2, true);
|
|
713
|
+
writeString(8, "WAVE");
|
|
714
|
+
writeString(12, "fmt ");
|
|
715
|
+
view.setUint32(16, 16, true);
|
|
716
|
+
view.setUint16(20, 1, true);
|
|
717
|
+
view.setUint16(22, channels, true);
|
|
718
|
+
view.setUint32(24, sampleRate, true);
|
|
719
|
+
view.setUint32(28, sampleRate * channels * 2, true);
|
|
720
|
+
view.setUint16(32, channels * 2, true);
|
|
721
|
+
view.setUint16(34, 16, true);
|
|
722
|
+
writeString(36, "data");
|
|
723
|
+
view.setUint32(40, samples.length * 2, true);
|
|
724
|
+
const dataOffset = 44;
|
|
725
|
+
for (let i = 0; i < samples.length; i++) {
|
|
726
|
+
view.setInt16(dataOffset + i * 2, samples[i] ?? 0, true);
|
|
727
|
+
}
|
|
728
|
+
return buffer;
|
|
729
|
+
},
|
|
730
|
+
/**
|
|
731
|
+
* ArrayBuffer 转 Base64
|
|
732
|
+
* @param buffer ArrayBuffer
|
|
733
|
+
* @returns Base64 字符串
|
|
734
|
+
*/
|
|
735
|
+
arrayBufferToBase64(buffer) {
|
|
736
|
+
let binary = "";
|
|
737
|
+
const bytes = new Uint8Array(buffer);
|
|
738
|
+
const len = bytes.byteLength;
|
|
739
|
+
for (let i = 0; i < len; i++) {
|
|
740
|
+
binary += String.fromCharCode(bytes[i] ?? 0);
|
|
741
|
+
}
|
|
742
|
+
return typeof btoa !== "undefined" ? btoa(binary) : Buffer.from(binary).toString("base64");
|
|
743
|
+
}
|
|
744
|
+
};
|
|
745
|
+
var WORKLET_CODE = `
|
|
746
|
+
class SpeechProcessor extends AudioWorkletProcessor {
|
|
747
|
+
constructor() {
|
|
748
|
+
super();
|
|
749
|
+
this.buffer = [];
|
|
750
|
+
this.targetRate = 16000;
|
|
751
|
+
this.currentRate = 44100;
|
|
752
|
+
this.silenceFrames = 0;
|
|
753
|
+
this.maxSilenceFrames = 0;
|
|
754
|
+
this.vadThreshold = 0.02;
|
|
755
|
+
this.isRecording = false;
|
|
756
|
+
this.port.onmessage = this.handleMessage.bind(this);
|
|
757
|
+
}
|
|
758
|
+
|
|
759
|
+
static get parameterDescriptors() {
|
|
760
|
+
return [];
|
|
761
|
+
}
|
|
762
|
+
|
|
763
|
+
process(inputs) {
|
|
764
|
+
const input = inputs[0];
|
|
765
|
+
if (!input || !input.length || !this.isRecording) return true;
|
|
766
|
+
const channelData = input[0];
|
|
767
|
+
|
|
768
|
+
// VAD \u68C0\u6D4B
|
|
769
|
+
let sum = 0;
|
|
770
|
+
for (let i = 0; i < channelData.length; i++) {
|
|
771
|
+
sum += channelData[i] * channelData[i];
|
|
772
|
+
}
|
|
773
|
+
const rms = Math.sqrt(sum / channelData.length);
|
|
774
|
+
|
|
775
|
+
if (rms < this.vadThreshold) {
|
|
776
|
+
this.silenceFrames++;
|
|
777
|
+
if (this.maxSilenceFrames > 0 && this.silenceFrames > this.maxSilenceFrames) {
|
|
778
|
+
this.port.postMessage({ type: 'VAD_TIMEOUT' });
|
|
779
|
+
this.silenceFrames = 0;
|
|
780
|
+
}
|
|
781
|
+
} else {
|
|
782
|
+
this.silenceFrames = 0;
|
|
783
|
+
}
|
|
784
|
+
|
|
785
|
+
// \u91CD\u91C7\u6837
|
|
786
|
+
if (this.targetRate < this.currentRate) {
|
|
787
|
+
const compression = this.currentRate / this.targetRate;
|
|
788
|
+
for (let i = 0; i < channelData.length; i += compression) {
|
|
789
|
+
this.buffer.push(channelData[Math.floor(i)]);
|
|
790
|
+
}
|
|
791
|
+
} else {
|
|
792
|
+
this.buffer.push(...channelData);
|
|
793
|
+
}
|
|
794
|
+
|
|
795
|
+
// \u5B9A\u671F\u5237\u65B0 (~128ms)
|
|
796
|
+
if (this.buffer.length >= 2048) {
|
|
797
|
+
this.flush();
|
|
798
|
+
}
|
|
799
|
+
return true;
|
|
800
|
+
}
|
|
801
|
+
|
|
802
|
+
flush() {
|
|
803
|
+
if (this.buffer.length === 0) return;
|
|
804
|
+
const pcmData = new Int16Array(this.buffer.length);
|
|
805
|
+
for (let i = 0; i < this.buffer.length; i++) {
|
|
806
|
+
let s = Math.max(-1, Math.min(1, this.buffer[i]));
|
|
807
|
+
s = s < 0 ? s * 0x8000 : s * 0x7FFF;
|
|
808
|
+
pcmData[i] = s;
|
|
809
|
+
}
|
|
810
|
+
this.port.postMessage({ type: 'AUDIO_DATA', payload: pcmData.buffer }, [pcmData.buffer]);
|
|
811
|
+
this.buffer = [];
|
|
812
|
+
}
|
|
813
|
+
|
|
814
|
+
handleMessage(event) {
|
|
815
|
+
const { type, payload } = event.data;
|
|
816
|
+
if (type === 'CONFIG') {
|
|
817
|
+
this.targetRate = payload.targetRate || 16000;
|
|
818
|
+
this.currentRate = payload.currentRate || 44100;
|
|
819
|
+
this.vadThreshold = payload.vadThreshold || 0.02;
|
|
820
|
+
const secondsPerBlock = 128 / this.currentRate;
|
|
821
|
+
this.maxSilenceFrames = (payload.vadDuration / 1000) / secondsPerBlock;
|
|
822
|
+
}
|
|
823
|
+
if (type === 'SET_RECORDING') {
|
|
824
|
+
this.isRecording = payload;
|
|
825
|
+
if (!payload) this.flush();
|
|
826
|
+
}
|
|
827
|
+
}
|
|
828
|
+
}
|
|
829
|
+
registerProcessor('speech-processor', SpeechProcessor);
|
|
830
|
+
`;
|
|
831
|
+
var BaseRecognitionStrategy = class {
|
|
832
|
+
constructor(config) {
|
|
833
|
+
this.config = config;
|
|
834
|
+
}
|
|
835
|
+
static {
|
|
836
|
+
chunkPK6SKIKE_cjs.__name(this, "BaseRecognitionStrategy");
|
|
837
|
+
}
|
|
838
|
+
listeners = {
|
|
839
|
+
result: [],
|
|
840
|
+
error: [],
|
|
841
|
+
state: [],
|
|
842
|
+
start: [],
|
|
843
|
+
end: [],
|
|
844
|
+
soundstart: [],
|
|
845
|
+
soundend: [],
|
|
846
|
+
speechstart: [],
|
|
847
|
+
speechend: [],
|
|
848
|
+
audiostart: [],
|
|
849
|
+
audioend: []
|
|
850
|
+
};
|
|
851
|
+
_status = "IDLE" /* IDLE */;
|
|
852
|
+
/**
|
|
853
|
+
* 获取当前状态
|
|
854
|
+
*/
|
|
855
|
+
get status() {
|
|
856
|
+
return this._status;
|
|
857
|
+
}
|
|
858
|
+
/**
|
|
859
|
+
* 设置状态
|
|
860
|
+
*/
|
|
861
|
+
setStatus(status) {
|
|
862
|
+
if (this._status !== status) {
|
|
863
|
+
this._status = status;
|
|
864
|
+
this.emit("state", status);
|
|
865
|
+
}
|
|
866
|
+
}
|
|
867
|
+
/**
|
|
868
|
+
* 添加事件监听
|
|
869
|
+
*/
|
|
870
|
+
on(event, fn) {
|
|
871
|
+
this.listeners[event].push(fn);
|
|
872
|
+
}
|
|
873
|
+
/**
|
|
874
|
+
* 移除事件监听
|
|
875
|
+
*/
|
|
876
|
+
off(event, fn) {
|
|
877
|
+
const listeners = this.listeners[event];
|
|
878
|
+
const index = listeners.indexOf(fn);
|
|
879
|
+
if (index > -1) {
|
|
880
|
+
listeners.splice(index, 1);
|
|
881
|
+
}
|
|
882
|
+
}
|
|
883
|
+
/**
|
|
884
|
+
* 触发事件
|
|
885
|
+
*/
|
|
886
|
+
emit(event, data) {
|
|
887
|
+
const listeners = this.listeners[event];
|
|
888
|
+
listeners.forEach((fn) => {
|
|
889
|
+
try {
|
|
890
|
+
fn(data);
|
|
891
|
+
} catch (e) {
|
|
892
|
+
console.error(`[Recognition] \u4E8B\u4EF6\u5904\u7406\u5668\u9519\u8BEF (${event}):`, e);
|
|
893
|
+
}
|
|
894
|
+
});
|
|
895
|
+
}
|
|
896
|
+
/**
|
|
897
|
+
* 销毁资源
|
|
898
|
+
*/
|
|
899
|
+
dispose() {
|
|
900
|
+
this.abort();
|
|
901
|
+
Object.keys(this.listeners).forEach((key) => {
|
|
902
|
+
this.listeners[key] = [];
|
|
903
|
+
});
|
|
904
|
+
}
|
|
905
|
+
};
|
|
906
|
+
var NativeRecognitionStrategy = class extends BaseRecognitionStrategy {
|
|
907
|
+
static {
|
|
908
|
+
chunkPK6SKIKE_cjs.__name(this, "NativeRecognitionStrategy");
|
|
909
|
+
}
|
|
910
|
+
recognition = null;
|
|
911
|
+
SpeechRecognitionClass = null;
|
|
912
|
+
constructor(config) {
|
|
913
|
+
super(config);
|
|
914
|
+
const srClass = typeof window !== "undefined" ? window.SpeechRecognition ?? window.webkitSpeechRecognition : void 0;
|
|
915
|
+
this.SpeechRecognitionClass = srClass ?? null;
|
|
916
|
+
}
|
|
917
|
+
/**
|
|
918
|
+
* 检查是否可用
|
|
919
|
+
*/
|
|
920
|
+
isAvailable() {
|
|
921
|
+
return this.SpeechRecognitionClass !== null;
|
|
922
|
+
}
|
|
923
|
+
async start() {
|
|
924
|
+
if (!this.SpeechRecognitionClass) {
|
|
925
|
+
throw { code: "NOT_SUPPORTED", message: "\u6D4F\u89C8\u5668\u4E0D\u652F\u6301\u8BED\u97F3\u8BC6\u522B" };
|
|
926
|
+
}
|
|
927
|
+
if (this._status !== "IDLE" /* IDLE */) {
|
|
928
|
+
return;
|
|
929
|
+
}
|
|
930
|
+
this.setStatus("CONNECTING" /* CONNECTING */);
|
|
931
|
+
return new Promise((resolve, reject) => {
|
|
932
|
+
this.recognition = new this.SpeechRecognitionClass();
|
|
933
|
+
this.recognition.lang = this.config.lang ?? "zh-CN";
|
|
934
|
+
this.recognition.continuous = this.config.continuous ?? false;
|
|
935
|
+
this.recognition.interimResults = this.config.interimResults ?? true;
|
|
936
|
+
this.recognition.maxAlternatives = this.config.maxAlternatives ?? 1;
|
|
937
|
+
this.recognition.onstart = () => {
|
|
938
|
+
this.setStatus("RECORDING" /* RECORDING */);
|
|
939
|
+
this.emit("start");
|
|
940
|
+
this.emit("audiostart");
|
|
941
|
+
resolve();
|
|
942
|
+
};
|
|
943
|
+
this.recognition.onend = () => {
|
|
944
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
945
|
+
this.emit("audioend");
|
|
946
|
+
this.emit("end");
|
|
947
|
+
};
|
|
948
|
+
this.recognition.onresult = (e) => {
|
|
949
|
+
const results = e.results[e.resultIndex];
|
|
950
|
+
if (results?.[0]) {
|
|
951
|
+
this.emit("result", {
|
|
952
|
+
transcript: results[0].transcript,
|
|
953
|
+
isFinal: results.isFinal,
|
|
954
|
+
confidence: results[0].confidence,
|
|
955
|
+
original: e
|
|
956
|
+
});
|
|
957
|
+
}
|
|
958
|
+
};
|
|
959
|
+
this.recognition.onerror = (e) => {
|
|
960
|
+
const errorMap = {
|
|
961
|
+
"not-allowed": "NOT_ALLOWED",
|
|
962
|
+
"no-speech": "NO_SPEECH",
|
|
963
|
+
network: "NETWORK"
|
|
964
|
+
};
|
|
965
|
+
const error = {
|
|
966
|
+
code: errorMap[e.error] ?? "UNKNOWN",
|
|
967
|
+
message: this.getErrorMessage(e.error),
|
|
968
|
+
originalError: e
|
|
969
|
+
};
|
|
970
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
971
|
+
this.emit("error", error);
|
|
972
|
+
if (e.error === "not-allowed" || e.error === "audio-capture") {
|
|
973
|
+
reject(error);
|
|
974
|
+
}
|
|
975
|
+
};
|
|
976
|
+
this.recognition.onsoundstart = () => this.emit("soundstart");
|
|
977
|
+
this.recognition.onsoundend = () => this.emit("soundend");
|
|
978
|
+
this.recognition.onspeechstart = () => this.emit("speechstart");
|
|
979
|
+
this.recognition.onspeechend = () => this.emit("speechend");
|
|
980
|
+
try {
|
|
981
|
+
this.recognition.start();
|
|
982
|
+
} catch (e) {
|
|
983
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
984
|
+
reject({ code: "UNKNOWN", message: "\u542F\u52A8\u8BC6\u522B\u5931\u8D25", originalError: e });
|
|
985
|
+
}
|
|
986
|
+
});
|
|
987
|
+
}
|
|
988
|
+
stop() {
|
|
989
|
+
if (this.recognition && this._status === "RECORDING" /* RECORDING */) {
|
|
990
|
+
this.recognition.stop();
|
|
991
|
+
}
|
|
992
|
+
}
|
|
993
|
+
abort() {
|
|
994
|
+
if (this.recognition) {
|
|
995
|
+
this.recognition.abort();
|
|
996
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
997
|
+
}
|
|
998
|
+
}
|
|
999
|
+
isListening() {
|
|
1000
|
+
return this._status === "RECORDING" /* RECORDING */;
|
|
1001
|
+
}
|
|
1002
|
+
getErrorMessage(errorCode) {
|
|
1003
|
+
const messages = {
|
|
1004
|
+
"no-speech": "\u672A\u68C0\u6D4B\u5230\u8BED\u97F3\u8F93\u5165",
|
|
1005
|
+
"audio-capture": "\u65E0\u6CD5\u8BBF\u95EE\u9EA6\u514B\u98CE",
|
|
1006
|
+
"not-allowed": "\u9EA6\u514B\u98CE\u6743\u9650\u88AB\u62D2\u7EDD",
|
|
1007
|
+
network: "\u7F51\u7EDC\u9519\u8BEF",
|
|
1008
|
+
aborted: "\u8BC6\u522B\u88AB\u4E2D\u6B62",
|
|
1009
|
+
"language-not-supported": "\u4E0D\u652F\u6301\u7684\u8BED\u8A00",
|
|
1010
|
+
"service-not-allowed": "\u8BED\u97F3\u670D\u52A1\u4E0D\u53EF\u7528"
|
|
1011
|
+
};
|
|
1012
|
+
return messages[errorCode] ?? `\u8BED\u97F3\u8BC6\u522B\u9519\u8BEF: ${errorCode}`;
|
|
1013
|
+
}
|
|
1014
|
+
};
|
|
1015
|
+
var CloudRecognitionStrategy = class extends BaseRecognitionStrategy {
|
|
1016
|
+
static {
|
|
1017
|
+
chunkPK6SKIKE_cjs.__name(this, "CloudRecognitionStrategy");
|
|
1018
|
+
}
|
|
1019
|
+
audioContext = null;
|
|
1020
|
+
mediaStream = null;
|
|
1021
|
+
workletNode = null;
|
|
1022
|
+
scriptProcessor = null;
|
|
1023
|
+
socket = null;
|
|
1024
|
+
adapter;
|
|
1025
|
+
transport;
|
|
1026
|
+
// HTTP 模式缓冲
|
|
1027
|
+
pcmChunks = [];
|
|
1028
|
+
totalPCMLength = 0;
|
|
1029
|
+
// 页面可见性监听器
|
|
1030
|
+
visibilityHandler = null;
|
|
1031
|
+
// 弱网消息队列
|
|
1032
|
+
msgQueue = [];
|
|
1033
|
+
MSG_QUEUE_MAX = 50;
|
|
1034
|
+
// ScriptProcessor 降级参数
|
|
1035
|
+
spSilenceCount = 0;
|
|
1036
|
+
spMaxSilence = 0;
|
|
1037
|
+
isRecordingFlag = false;
|
|
1038
|
+
// 重连参数
|
|
1039
|
+
reconnectAttempts = 0;
|
|
1040
|
+
reconnectTimer = null;
|
|
1041
|
+
// Worklet 加载标记
|
|
1042
|
+
workletLoaded = false;
|
|
1043
|
+
constructor(config) {
|
|
1044
|
+
super(config);
|
|
1045
|
+
if (!config.cloudAdapter) {
|
|
1046
|
+
throw new Error("[CloudStrategy] \u9700\u8981\u63D0\u4F9B cloudAdapter");
|
|
1047
|
+
}
|
|
1048
|
+
this.adapter = config.cloudAdapter;
|
|
1049
|
+
this.transport = config.transport ?? "websocket";
|
|
1050
|
+
}
|
|
1051
|
+
async start() {
|
|
1052
|
+
if (this._status !== "IDLE" /* IDLE */) {
|
|
1053
|
+
return;
|
|
1054
|
+
}
|
|
1055
|
+
this.setStatus("CONNECTING" /* CONNECTING */);
|
|
1056
|
+
this.pcmChunks = [];
|
|
1057
|
+
this.totalPCMLength = 0;
|
|
1058
|
+
this.reconnectAttempts = 0;
|
|
1059
|
+
try {
|
|
1060
|
+
if (this.transport === "websocket") {
|
|
1061
|
+
if (!this.adapter.getConnectUrl) {
|
|
1062
|
+
throw new Error("\u9002\u914D\u5668\u7F3A\u5C11 getConnectUrl \u65B9\u6CD5");
|
|
1063
|
+
}
|
|
1064
|
+
const url = await Promise.resolve(this.adapter.getConnectUrl());
|
|
1065
|
+
await this.initWebSocket(url);
|
|
1066
|
+
}
|
|
1067
|
+
const audioConfig = this.config.audioConfig ?? {};
|
|
1068
|
+
this.mediaStream = await navigator.mediaDevices.getUserMedia({
|
|
1069
|
+
audio: {
|
|
1070
|
+
echoCancellation: audioConfig.echoCancellation ?? true,
|
|
1071
|
+
noiseSuppression: audioConfig.noiseSuppression ?? true,
|
|
1072
|
+
autoGainControl: audioConfig.autoGainControl ?? true
|
|
1073
|
+
}
|
|
1074
|
+
});
|
|
1075
|
+
await this.initAudioContext();
|
|
1076
|
+
if (this.transport === "http") {
|
|
1077
|
+
this.setStatus("RECORDING" /* RECORDING */);
|
|
1078
|
+
}
|
|
1079
|
+
this.setupVisibilityListener();
|
|
1080
|
+
this.emit("start");
|
|
1081
|
+
this.emit("audiostart");
|
|
1082
|
+
} catch (err) {
|
|
1083
|
+
const error = err;
|
|
1084
|
+
this.emit("error", {
|
|
1085
|
+
code: "NOT_ALLOWED",
|
|
1086
|
+
message: error.message || "\u542F\u52A8\u5931\u8D25",
|
|
1087
|
+
originalError: err
|
|
1088
|
+
});
|
|
1089
|
+
this.cleanup();
|
|
1090
|
+
throw err;
|
|
1091
|
+
}
|
|
1092
|
+
}
|
|
1093
|
+
/**
|
|
1094
|
+
* 初始化 WebSocket 连接
|
|
1095
|
+
*/
|
|
1096
|
+
initWebSocket(url) {
|
|
1097
|
+
return new Promise((resolve, reject) => {
|
|
1098
|
+
this.socket = new WebSocket(url);
|
|
1099
|
+
this.socket.binaryType = "arraybuffer";
|
|
1100
|
+
const timeout = setTimeout(() => {
|
|
1101
|
+
reject(new Error("WebSocket \u8FDE\u63A5\u8D85\u65F6"));
|
|
1102
|
+
}, 1e4);
|
|
1103
|
+
this.socket.onopen = () => {
|
|
1104
|
+
clearTimeout(timeout);
|
|
1105
|
+
this.setStatus("RECORDING" /* RECORDING */);
|
|
1106
|
+
this.reconnectAttempts = 0;
|
|
1107
|
+
if (this.adapter.getHandshakeParams) {
|
|
1108
|
+
const params = this.adapter.getHandshakeParams();
|
|
1109
|
+
if (params) {
|
|
1110
|
+
const msg = typeof params === "string" ? params : JSON.stringify(params);
|
|
1111
|
+
this.socket?.send(msg);
|
|
1112
|
+
}
|
|
1113
|
+
}
|
|
1114
|
+
resolve();
|
|
1115
|
+
};
|
|
1116
|
+
this.socket.onerror = (e) => {
|
|
1117
|
+
clearTimeout(timeout);
|
|
1118
|
+
reject(e);
|
|
1119
|
+
};
|
|
1120
|
+
this.socket.onmessage = (e) => {
|
|
1121
|
+
try {
|
|
1122
|
+
const data = e.data;
|
|
1123
|
+
const raw = typeof data === "string" ? data : new TextDecoder().decode(data);
|
|
1124
|
+
const res = this.adapter.parseResult(JSON.parse(raw));
|
|
1125
|
+
if (res) {
|
|
1126
|
+
this.emit("result", res);
|
|
1127
|
+
}
|
|
1128
|
+
} catch {
|
|
1129
|
+
}
|
|
1130
|
+
};
|
|
1131
|
+
this.socket.onclose = () => {
|
|
1132
|
+
if (this._status === "RECORDING" /* RECORDING */) {
|
|
1133
|
+
this.handleReconnect();
|
|
1134
|
+
}
|
|
1135
|
+
};
|
|
1136
|
+
});
|
|
1137
|
+
}
|
|
1138
|
+
/**
|
|
1139
|
+
* 处理重连逻辑
|
|
1140
|
+
*/
|
|
1141
|
+
handleReconnect() {
|
|
1142
|
+
const maxAttempts = this.config.maxReconnectAttempts ?? 3;
|
|
1143
|
+
const interval = this.config.reconnectInterval ?? 2e3;
|
|
1144
|
+
if (this.config.autoReconnect && this.reconnectAttempts < maxAttempts) {
|
|
1145
|
+
this.reconnectAttempts++;
|
|
1146
|
+
console.log(`[CloudStrategy] \u5C1D\u8BD5\u91CD\u8FDE (${this.reconnectAttempts}/${maxAttempts})`);
|
|
1147
|
+
this.reconnectTimer = setTimeout(() => {
|
|
1148
|
+
void (async () => {
|
|
1149
|
+
try {
|
|
1150
|
+
if (this.adapter.getConnectUrl) {
|
|
1151
|
+
const url = await Promise.resolve(this.adapter.getConnectUrl());
|
|
1152
|
+
await this.initWebSocket(url);
|
|
1153
|
+
}
|
|
1154
|
+
} catch {
|
|
1155
|
+
this.handleReconnect();
|
|
1156
|
+
}
|
|
1157
|
+
})();
|
|
1158
|
+
}, interval);
|
|
1159
|
+
} else {
|
|
1160
|
+
this.emit("error", {
|
|
1161
|
+
code: "NETWORK",
|
|
1162
|
+
message: "\u8FDE\u63A5\u5DF2\u65AD\u5F00"
|
|
1163
|
+
});
|
|
1164
|
+
void this.stop();
|
|
1165
|
+
}
|
|
1166
|
+
}
|
|
1167
|
+
/**
|
|
1168
|
+
* 初始化音频上下文
|
|
1169
|
+
*/
|
|
1170
|
+
async initAudioContext() {
|
|
1171
|
+
const AudioCtor = typeof window !== "undefined" && (window.AudioContext || window.webkitAudioContext);
|
|
1172
|
+
if (!AudioCtor) {
|
|
1173
|
+
throw new Error("\u6D4F\u89C8\u5668\u4E0D\u652F\u6301 AudioContext");
|
|
1174
|
+
}
|
|
1175
|
+
this.audioContext = new AudioCtor();
|
|
1176
|
+
if (this.audioContext.state === "suspended") {
|
|
1177
|
+
await this.audioContext.resume();
|
|
1178
|
+
}
|
|
1179
|
+
const source = this.audioContext.createMediaStreamSource(this.mediaStream);
|
|
1180
|
+
if (this.audioContext.audioWorklet && !this.workletLoaded) {
|
|
1181
|
+
try {
|
|
1182
|
+
const blob = new Blob([WORKLET_CODE], { type: "application/javascript" });
|
|
1183
|
+
await this.audioContext.audioWorklet.addModule(URL.createObjectURL(blob));
|
|
1184
|
+
this.workletLoaded = true;
|
|
1185
|
+
this.setupWorklet(source);
|
|
1186
|
+
console.log("[CloudStrategy] \u4F7F\u7528 AudioWorklet \u6A21\u5F0F");
|
|
1187
|
+
return;
|
|
1188
|
+
} catch (e) {
|
|
1189
|
+
console.warn("[CloudStrategy] AudioWorklet \u52A0\u8F7D\u5931\u8D25\uFF0C\u964D\u7EA7\u5230 ScriptProcessor:", e);
|
|
1190
|
+
}
|
|
1191
|
+
}
|
|
1192
|
+
this.setupScriptProcessor(source);
|
|
1193
|
+
console.log("[CloudStrategy] \u4F7F\u7528 ScriptProcessor \u964D\u7EA7\u6A21\u5F0F");
|
|
1194
|
+
}
|
|
1195
|
+
/**
|
|
1196
|
+
* 设置 AudioWorklet
|
|
1197
|
+
*/
|
|
1198
|
+
setupWorklet(source) {
|
|
1199
|
+
this.workletNode = new AudioWorkletNode(this.audioContext, "speech-processor");
|
|
1200
|
+
const audioConfig = this.config.audioConfig ?? {};
|
|
1201
|
+
this.workletNode.port.postMessage({
|
|
1202
|
+
type: "CONFIG",
|
|
1203
|
+
payload: {
|
|
1204
|
+
currentRate: this.audioContext.sampleRate,
|
|
1205
|
+
targetRate: audioConfig.sampleRate ?? 16e3,
|
|
1206
|
+
vadThreshold: audioConfig.vadThreshold ?? 0.02,
|
|
1207
|
+
vadDuration: audioConfig.vadDuration ?? 3e3
|
|
1208
|
+
}
|
|
1209
|
+
});
|
|
1210
|
+
this.workletNode.port.onmessage = (e) => {
|
|
1211
|
+
const { type, payload } = e.data;
|
|
1212
|
+
if (type === "AUDIO_DATA" && payload) {
|
|
1213
|
+
this.handlePCM(payload);
|
|
1214
|
+
}
|
|
1215
|
+
if (type === "VAD_TIMEOUT") {
|
|
1216
|
+
this.emit("speechend");
|
|
1217
|
+
void this.stop();
|
|
1218
|
+
}
|
|
1219
|
+
};
|
|
1220
|
+
this.workletNode.port.postMessage({ type: "SET_RECORDING", payload: true });
|
|
1221
|
+
source.connect(this.workletNode);
|
|
1222
|
+
this.workletNode.connect(this.audioContext.destination);
|
|
1223
|
+
this.emit("speechstart");
|
|
1224
|
+
}
|
|
1225
|
+
/**
|
|
1226
|
+
* 设置 ScriptProcessor (降级方案)
|
|
1227
|
+
*/
|
|
1228
|
+
setupScriptProcessor(source) {
|
|
1229
|
+
const bufferSize = 4096;
|
|
1230
|
+
this.scriptProcessor = this.audioContext.createScriptProcessor(bufferSize, 1, 1);
|
|
1231
|
+
const audioConfig = this.config.audioConfig ?? {};
|
|
1232
|
+
const targetRate = audioConfig.sampleRate ?? 16e3;
|
|
1233
|
+
const currentRate = this.audioContext.sampleRate;
|
|
1234
|
+
const secondsPerBuf = bufferSize / currentRate;
|
|
1235
|
+
this.spMaxSilence = (audioConfig.vadDuration ?? 3e3) / 1e3 / secondsPerBuf;
|
|
1236
|
+
this.spSilenceCount = 0;
|
|
1237
|
+
this.isRecordingFlag = true;
|
|
1238
|
+
this.scriptProcessor.onaudioprocess = (e) => {
|
|
1239
|
+
if (!this.isRecordingFlag) return;
|
|
1240
|
+
const input = e.inputBuffer.getChannelData(0);
|
|
1241
|
+
const rms = AudioUtils.calculateRMS(input);
|
|
1242
|
+
if (rms < (audioConfig.vadThreshold ?? 0.02)) {
|
|
1243
|
+
this.spSilenceCount++;
|
|
1244
|
+
if (this.spSilenceCount > this.spMaxSilence) {
|
|
1245
|
+
this.emit("speechend");
|
|
1246
|
+
void this.stop();
|
|
1247
|
+
return;
|
|
1248
|
+
}
|
|
1249
|
+
} else {
|
|
1250
|
+
this.spSilenceCount = 0;
|
|
1251
|
+
}
|
|
1252
|
+
const resampled = AudioUtils.resample(input, currentRate, targetRate);
|
|
1253
|
+
const pcm = AudioUtils.floatTo16BitPCM(resampled);
|
|
1254
|
+
this.handlePCM(pcm.buffer.slice(0));
|
|
1255
|
+
};
|
|
1256
|
+
source.connect(this.scriptProcessor);
|
|
1257
|
+
this.scriptProcessor.connect(this.audioContext.destination);
|
|
1258
|
+
this.emit("speechstart");
|
|
1259
|
+
}
|
|
1260
|
+
/**
|
|
1261
|
+
* 处理 PCM 数据
|
|
1262
|
+
*/
|
|
1263
|
+
handlePCM(buffer) {
|
|
1264
|
+
if (this.transport === "websocket") {
|
|
1265
|
+
let payload = buffer;
|
|
1266
|
+
if (this.adapter.transformAudioData) {
|
|
1267
|
+
payload = this.adapter.transformAudioData(buffer);
|
|
1268
|
+
}
|
|
1269
|
+
if (this.socket && this.socket.readyState === WebSocket.OPEN) {
|
|
1270
|
+
this.flushMsgQueue();
|
|
1271
|
+
this.socket.send(payload);
|
|
1272
|
+
} else if (this.socket && this.socket.readyState === WebSocket.CONNECTING) {
|
|
1273
|
+
if (payload instanceof ArrayBuffer && this.msgQueue.length < this.MSG_QUEUE_MAX) {
|
|
1274
|
+
this.msgQueue.push(payload);
|
|
1275
|
+
}
|
|
1276
|
+
}
|
|
1277
|
+
} else {
|
|
1278
|
+
const chunk = buffer.slice(0);
|
|
1279
|
+
this.pcmChunks.push(chunk);
|
|
1280
|
+
this.totalPCMLength += chunk.byteLength / 2;
|
|
1281
|
+
}
|
|
1282
|
+
}
|
|
1283
|
+
/**
|
|
1284
|
+
* 刷新消息队列
|
|
1285
|
+
*/
|
|
1286
|
+
flushMsgQueue() {
|
|
1287
|
+
while (this.msgQueue.length > 0 && this.socket?.readyState === WebSocket.OPEN) {
|
|
1288
|
+
const data = this.msgQueue.shift();
|
|
1289
|
+
if (data) {
|
|
1290
|
+
this.socket.send(data);
|
|
1291
|
+
}
|
|
1292
|
+
}
|
|
1293
|
+
}
|
|
1294
|
+
/**
|
|
1295
|
+
* 设置页面可见性监听
|
|
1296
|
+
*/
|
|
1297
|
+
setupVisibilityListener() {
|
|
1298
|
+
if (typeof document === "undefined") return;
|
|
1299
|
+
this.visibilityHandler = () => {
|
|
1300
|
+
if (document.hidden && this._status === "RECORDING" /* RECORDING */) {
|
|
1301
|
+
this.emit("error", {
|
|
1302
|
+
code: "NOT_ALLOWED",
|
|
1303
|
+
message: "\u9875\u9762\u5DF2\u8FDB\u5165\u540E\u53F0\uFF0C\u5F55\u97F3\u5DF2\u6682\u505C"
|
|
1304
|
+
});
|
|
1305
|
+
void this.stop();
|
|
1306
|
+
}
|
|
1307
|
+
};
|
|
1308
|
+
document.addEventListener("visibilitychange", this.visibilityHandler);
|
|
1309
|
+
}
|
|
1310
|
+
async stop() {
|
|
1311
|
+
if (this._status === "IDLE" /* IDLE */) {
|
|
1312
|
+
return;
|
|
1313
|
+
}
|
|
1314
|
+
this.isRecordingFlag = false;
|
|
1315
|
+
this.workletNode?.port.postMessage({ type: "SET_RECORDING", payload: false });
|
|
1316
|
+
this.mediaStream?.getTracks().forEach((t) => t.stop());
|
|
1317
|
+
if (this.transport === "http" && this.pcmChunks.length > 0 && this.adapter.recognizeShortAudio) {
|
|
1318
|
+
try {
|
|
1319
|
+
this.setStatus("PROCESSING" /* PROCESSING */);
|
|
1320
|
+
const mergedPCM = AudioUtils.mergeBuffers(this.pcmChunks, this.totalPCMLength);
|
|
1321
|
+
const sampleRate = this.config.audioConfig?.sampleRate ?? 16e3;
|
|
1322
|
+
const wavBuffer = AudioUtils.encodeWAV(mergedPCM, sampleRate);
|
|
1323
|
+
const result = await this.adapter.recognizeShortAudio(wavBuffer);
|
|
1324
|
+
this.emit("result", result);
|
|
1325
|
+
} catch (e) {
|
|
1326
|
+
const error = e;
|
|
1327
|
+
this.emit("error", {
|
|
1328
|
+
code: "ADAPTER_ERROR",
|
|
1329
|
+
message: error.message || "\u8BC6\u522B\u5931\u8D25",
|
|
1330
|
+
originalError: e
|
|
1331
|
+
});
|
|
1332
|
+
}
|
|
1333
|
+
}
|
|
1334
|
+
this.cleanup();
|
|
1335
|
+
this.emit("audioend");
|
|
1336
|
+
this.emit("end");
|
|
1337
|
+
}
|
|
1338
|
+
abort() {
|
|
1339
|
+
this.isRecordingFlag = false;
|
|
1340
|
+
this.cleanup();
|
|
1341
|
+
this.emit("end");
|
|
1342
|
+
}
|
|
1343
|
+
isListening() {
|
|
1344
|
+
return this._status === "RECORDING" /* RECORDING */;
|
|
1345
|
+
}
|
|
1346
|
+
/**
|
|
1347
|
+
* 清理资源
|
|
1348
|
+
*/
|
|
1349
|
+
cleanup() {
|
|
1350
|
+
if (this.reconnectTimer) {
|
|
1351
|
+
clearTimeout(this.reconnectTimer);
|
|
1352
|
+
this.reconnectTimer = null;
|
|
1353
|
+
}
|
|
1354
|
+
if (this.socket) {
|
|
1355
|
+
this.socket.close();
|
|
1356
|
+
this.socket = null;
|
|
1357
|
+
}
|
|
1358
|
+
this.workletNode?.disconnect();
|
|
1359
|
+
this.scriptProcessor?.disconnect();
|
|
1360
|
+
void this.audioContext?.close();
|
|
1361
|
+
if (this.visibilityHandler && typeof document !== "undefined") {
|
|
1362
|
+
document.removeEventListener("visibilitychange", this.visibilityHandler);
|
|
1363
|
+
this.visibilityHandler = null;
|
|
1364
|
+
}
|
|
1365
|
+
this.workletNode = null;
|
|
1366
|
+
this.scriptProcessor = null;
|
|
1367
|
+
this.audioContext = null;
|
|
1368
|
+
this.mediaStream = null;
|
|
1369
|
+
this.pcmChunks = [];
|
|
1370
|
+
this.totalPCMLength = 0;
|
|
1371
|
+
this.msgQueue = [];
|
|
1372
|
+
this.setStatus("IDLE" /* IDLE */);
|
|
1373
|
+
}
|
|
1374
|
+
};
|
|
1375
|
+
var SpeechRecognizerImpl = class {
|
|
1376
|
+
static {
|
|
1377
|
+
chunkPK6SKIKE_cjs.__name(this, "SpeechRecognizerImpl");
|
|
1378
|
+
}
|
|
1379
|
+
strategy = null;
|
|
1380
|
+
config = {};
|
|
1381
|
+
_currentProvider = "browser";
|
|
1382
|
+
_status = "idle";
|
|
1383
|
+
eventHandlers = /* @__PURE__ */ new Map();
|
|
1384
|
+
customProviders = /* @__PURE__ */ new Map();
|
|
1385
|
+
get currentProvider() {
|
|
1386
|
+
return this._currentProvider;
|
|
1387
|
+
}
|
|
1388
|
+
get status() {
|
|
1389
|
+
return this._status;
|
|
1390
|
+
}
|
|
1391
|
+
/**
|
|
1392
|
+
* 获取当前识别状态
|
|
1393
|
+
*/
|
|
1394
|
+
get recognitionStatus() {
|
|
1395
|
+
return this.strategy?.status ?? "IDLE" /* IDLE */;
|
|
1396
|
+
}
|
|
1397
|
+
/**
|
|
1398
|
+
* 初始化语音识别器
|
|
1399
|
+
*/
|
|
1400
|
+
initialize(config) {
|
|
1401
|
+
this._status = "loading";
|
|
1402
|
+
this.config = {
|
|
1403
|
+
lang: "zh-CN",
|
|
1404
|
+
continuous: false,
|
|
1405
|
+
interimResults: true,
|
|
1406
|
+
maxAlternatives: 1,
|
|
1407
|
+
mode: "auto",
|
|
1408
|
+
autoReconnect: true,
|
|
1409
|
+
maxReconnectAttempts: 3,
|
|
1410
|
+
reconnectInterval: 2e3,
|
|
1411
|
+
audioConfig: {
|
|
1412
|
+
sampleRate: 16e3,
|
|
1413
|
+
vadThreshold: 0.02,
|
|
1414
|
+
vadDuration: 3e3,
|
|
1415
|
+
echoCancellation: true,
|
|
1416
|
+
noiseSuppression: true,
|
|
1417
|
+
autoGainControl: true
|
|
1418
|
+
},
|
|
1419
|
+
...config
|
|
1420
|
+
};
|
|
1421
|
+
try {
|
|
1422
|
+
this.initializeStrategy();
|
|
1423
|
+
this._status = "ready";
|
|
1424
|
+
return Promise.resolve();
|
|
1425
|
+
} catch (error) {
|
|
1426
|
+
this._status = "error";
|
|
1427
|
+
return Promise.reject(error);
|
|
1428
|
+
}
|
|
1429
|
+
}
|
|
1430
|
+
/**
|
|
1431
|
+
* 初始化识别策略
|
|
1432
|
+
*/
|
|
1433
|
+
initializeStrategy() {
|
|
1434
|
+
const advConfig = this.config;
|
|
1435
|
+
const mode = advConfig.mode ?? "auto";
|
|
1436
|
+
const hasNative = typeof window !== "undefined" && ("SpeechRecognition" in window || "webkitSpeechRecognition" in window);
|
|
1437
|
+
if (mode === "native" || mode === "auto" && hasNative && !advConfig.cloudAdapter) {
|
|
1438
|
+
const nativeStrategy = new NativeRecognitionStrategy(advConfig);
|
|
1439
|
+
if (!nativeStrategy.isAvailable()) {
|
|
1440
|
+
throw new Error("\u6D4F\u89C8\u5668\u4E0D\u652F\u6301\u539F\u751F\u8BED\u97F3\u8BC6\u522B");
|
|
1441
|
+
}
|
|
1442
|
+
this.strategy = nativeStrategy;
|
|
1443
|
+
this._currentProvider = "browser";
|
|
1444
|
+
} else if (mode === "cloud" || mode === "auto" && advConfig.cloudAdapter) {
|
|
1445
|
+
if (!advConfig.cloudAdapter) {
|
|
1446
|
+
throw new Error("\u4E91\u7AEF\u6A21\u5F0F\u9700\u8981\u63D0\u4F9B cloudAdapter");
|
|
1447
|
+
}
|
|
1448
|
+
this.strategy = new CloudRecognitionStrategy(advConfig);
|
|
1449
|
+
this._currentProvider = "custom";
|
|
1450
|
+
} else {
|
|
1451
|
+
throw new Error("\u6CA1\u6709\u53EF\u7528\u7684\u8BC6\u522B\u7B56\u7565");
|
|
1452
|
+
}
|
|
1453
|
+
this.forwardStrategyEvents();
|
|
1454
|
+
}
|
|
1455
|
+
/**
|
|
1456
|
+
* 转发策略事件
|
|
1457
|
+
*/
|
|
1458
|
+
forwardStrategyEvents() {
|
|
1459
|
+
if (!this.strategy) return;
|
|
1460
|
+
this.strategy.on("result", (res) => {
|
|
1461
|
+
this.emit("result", {
|
|
1462
|
+
type: "result",
|
|
1463
|
+
result: this.convertResult(res)
|
|
1464
|
+
});
|
|
1465
|
+
});
|
|
1466
|
+
this.strategy.on("error", (err) => {
|
|
1467
|
+
const errorEvent = {
|
|
1468
|
+
type: "error",
|
|
1469
|
+
error: {
|
|
1470
|
+
code: err.code,
|
|
1471
|
+
message: err.message
|
|
1472
|
+
}
|
|
1473
|
+
};
|
|
1474
|
+
if (err.originalError) {
|
|
1475
|
+
errorEvent.error.originalError = err.originalError;
|
|
1476
|
+
}
|
|
1477
|
+
this.emit("error", errorEvent);
|
|
1478
|
+
});
|
|
1479
|
+
this.strategy.on("start", () => this.emit("start", { type: "start" }));
|
|
1480
|
+
this.strategy.on("end", () => this.emit("end", { type: "end" }));
|
|
1481
|
+
this.strategy.on("soundstart", () => this.emit("soundstart", { type: "soundstart" }));
|
|
1482
|
+
this.strategy.on("soundend", () => this.emit("soundend", { type: "soundend" }));
|
|
1483
|
+
this.strategy.on("speechstart", () => this.emit("speechstart", { type: "speechstart" }));
|
|
1484
|
+
this.strategy.on("speechend", () => this.emit("speechend", { type: "speechend" }));
|
|
1485
|
+
this.strategy.on("audiostart", () => this.emit("audiostart", { type: "audiostart" }));
|
|
1486
|
+
this.strategy.on("audioend", () => this.emit("audioend", { type: "audioend" }));
|
|
1487
|
+
}
|
|
1488
|
+
/**
|
|
1489
|
+
* 转换识别结果格式
|
|
1490
|
+
*/
|
|
1491
|
+
convertResult(res) {
|
|
1492
|
+
return {
|
|
1493
|
+
results: [
|
|
1494
|
+
{
|
|
1495
|
+
transcript: res.transcript,
|
|
1496
|
+
confidence: res.confidence,
|
|
1497
|
+
isFinal: res.isFinal
|
|
1498
|
+
}
|
|
1499
|
+
],
|
|
1500
|
+
bestTranscript: res.transcript,
|
|
1501
|
+
bestConfidence: res.confidence,
|
|
1502
|
+
isFinal: res.isFinal
|
|
1503
|
+
};
|
|
1504
|
+
}
|
|
1505
|
+
/**
|
|
1506
|
+
* 开始识别
|
|
1507
|
+
*/
|
|
1508
|
+
async start(config) {
|
|
1509
|
+
if (this._status !== "ready") {
|
|
1510
|
+
throw new Error("\u8BC6\u522B\u5668\u672A\u5C31\u7EEA");
|
|
1511
|
+
}
|
|
1512
|
+
if (config) {
|
|
1513
|
+
this.config = { ...this.config, ...config };
|
|
1514
|
+
this.initializeStrategy();
|
|
1515
|
+
}
|
|
1516
|
+
return this.strategy?.start();
|
|
1517
|
+
}
|
|
1518
|
+
/**
|
|
1519
|
+
* 停止识别
|
|
1520
|
+
*/
|
|
1521
|
+
stop() {
|
|
1522
|
+
this.strategy?.stop();
|
|
1523
|
+
}
|
|
1524
|
+
/**
|
|
1525
|
+
* 中止识别
|
|
1526
|
+
*/
|
|
1527
|
+
abort() {
|
|
1528
|
+
this.strategy?.abort();
|
|
1529
|
+
}
|
|
1530
|
+
/**
|
|
1531
|
+
* 是否正在识别
|
|
1532
|
+
*/
|
|
1533
|
+
isListening() {
|
|
1534
|
+
return this.strategy?.isListening() ?? false;
|
|
1535
|
+
}
|
|
1536
|
+
/**
|
|
1537
|
+
* 添加事件监听
|
|
1538
|
+
*/
|
|
1539
|
+
on(event, handler) {
|
|
1540
|
+
if (!this.eventHandlers.has(event)) {
|
|
1541
|
+
this.eventHandlers.set(event, /* @__PURE__ */ new Set());
|
|
1542
|
+
}
|
|
1543
|
+
this.eventHandlers.get(event).add(handler);
|
|
1544
|
+
}
|
|
1545
|
+
/**
|
|
1546
|
+
* 移除事件监听
|
|
1547
|
+
*/
|
|
1548
|
+
off(event, handler) {
|
|
1549
|
+
this.eventHandlers.get(event)?.delete(handler);
|
|
1550
|
+
}
|
|
1551
|
+
/**
|
|
1552
|
+
* 触发事件
|
|
1553
|
+
*/
|
|
1554
|
+
emit(type, event) {
|
|
1555
|
+
this.eventHandlers.get(type)?.forEach((handler) => {
|
|
1556
|
+
try {
|
|
1557
|
+
handler(event);
|
|
1558
|
+
} catch (e) {
|
|
1559
|
+
console.error("[SpeechRecognizer] \u4E8B\u4EF6\u5904\u7406\u5668\u9519\u8BEF:", e);
|
|
1560
|
+
}
|
|
1561
|
+
});
|
|
1562
|
+
}
|
|
1563
|
+
/**
|
|
1564
|
+
* 销毁实例
|
|
1565
|
+
*/
|
|
1566
|
+
dispose() {
|
|
1567
|
+
this.strategy?.dispose();
|
|
1568
|
+
this.strategy = null;
|
|
1569
|
+
this.eventHandlers.clear();
|
|
1570
|
+
this._status = "idle";
|
|
1571
|
+
}
|
|
1572
|
+
/**
|
|
1573
|
+
* 注册自定义提供商
|
|
1574
|
+
*/
|
|
1575
|
+
registerProvider(type, provider) {
|
|
1576
|
+
this.customProviders.set(type, provider);
|
|
1577
|
+
}
|
|
1578
|
+
/**
|
|
1579
|
+
* 使用云端适配器
|
|
1580
|
+
*/
|
|
1581
|
+
useCloudAdapter(adapter) {
|
|
1582
|
+
this.config.cloudAdapter = adapter;
|
|
1583
|
+
this.config.mode = "cloud";
|
|
1584
|
+
}
|
|
1585
|
+
};
|
|
1586
|
+
function getSpeechRecognitionConstructor() {
|
|
1587
|
+
if (typeof window === "undefined") {
|
|
1588
|
+
return null;
|
|
1589
|
+
}
|
|
1590
|
+
return window.SpeechRecognition ?? window.webkitSpeechRecognition ?? null;
|
|
1591
|
+
}
|
|
1592
|
+
chunkPK6SKIKE_cjs.__name(getSpeechRecognitionConstructor, "getSpeechRecognitionConstructor");
|
|
1593
|
+
async function createSpeechRecognizer(config) {
|
|
1594
|
+
const recognizer = new SpeechRecognizerImpl();
|
|
1595
|
+
await recognizer.initialize(config);
|
|
1596
|
+
return recognizer;
|
|
1597
|
+
}
|
|
1598
|
+
chunkPK6SKIKE_cjs.__name(createSpeechRecognizer, "createSpeechRecognizer");
|
|
1599
|
+
function isSpeechRecognitionSupported() {
|
|
1600
|
+
return getSpeechRecognitionConstructor() !== null;
|
|
1601
|
+
}
|
|
1602
|
+
chunkPK6SKIKE_cjs.__name(isSpeechRecognitionSupported, "isSpeechRecognitionSupported");
|
|
1603
|
+
async function listen(config) {
|
|
1604
|
+
const recognizer = await createSpeechRecognizer({
|
|
1605
|
+
...config,
|
|
1606
|
+
continuous: false
|
|
1607
|
+
});
|
|
1608
|
+
return new Promise((resolve, reject) => {
|
|
1609
|
+
let hasResult = false;
|
|
1610
|
+
recognizer.on("result", (event) => {
|
|
1611
|
+
if (event.result && event.result.isFinal) {
|
|
1612
|
+
hasResult = true;
|
|
1613
|
+
recognizer.dispose();
|
|
1614
|
+
resolve(event.result);
|
|
1615
|
+
}
|
|
1616
|
+
});
|
|
1617
|
+
recognizer.on("error", (event) => {
|
|
1618
|
+
recognizer.dispose();
|
|
1619
|
+
reject(event.error);
|
|
1620
|
+
});
|
|
1621
|
+
recognizer.on("end", () => {
|
|
1622
|
+
if (!hasResult) {
|
|
1623
|
+
recognizer.dispose();
|
|
1624
|
+
resolve({
|
|
1625
|
+
results: [],
|
|
1626
|
+
bestTranscript: "",
|
|
1627
|
+
bestConfidence: 0,
|
|
1628
|
+
isFinal: true
|
|
1629
|
+
});
|
|
1630
|
+
}
|
|
1631
|
+
});
|
|
1632
|
+
recognizer.start().catch(reject);
|
|
1633
|
+
});
|
|
1634
|
+
}
|
|
1635
|
+
chunkPK6SKIKE_cjs.__name(listen, "listen");
|
|
1636
|
+
|
|
1637
|
+
exports.SpeechRecognizerImpl = SpeechRecognizerImpl;
|
|
1638
|
+
exports.SpeechSynthesizerImpl = SpeechSynthesizerImpl;
|
|
1639
|
+
exports.createSpeechRecognizer = createSpeechRecognizer;
|
|
1640
|
+
exports.createSpeechSynthesizer = createSpeechSynthesizer;
|
|
1641
|
+
exports.isSpeechRecognitionSupported = isSpeechRecognitionSupported;
|
|
1642
|
+
exports.isSpeechSynthesisSupported = isSpeechSynthesisSupported;
|
|
1643
|
+
exports.listen = listen;
|
|
1644
|
+
exports.speak = speak;
|
|
1645
|
+
//# sourceMappingURL=chunk-YZVCK6VZ.cjs.map
|
|
1646
|
+
//# sourceMappingURL=chunk-YZVCK6VZ.cjs.map
|