fix: 优化模块

src/tts/provider/volcengine/test/mix.ts

@@ -1,5 +1,5 @@
 import { config } from './common.ts';
-import { runDemo } from '../tts-mix.ts';
+import { runDemo, TtsMix } from '../tts-mix.ts';
 const appId = config.APP_ID;
 const token = config.TOKEN;
 
@@ -7,11 +7,50 @@ const token = config.TOKEN;
 const speaker = 'zh_female_roumeinvyou_emo_v2_mars_bigtts';
 const text = '明朝开国皇帝朱元璋也称这本书为,万物之根';
 const outputPath = 'videos/tts_mix.wav';
-
+const text2 =
+  '明朝开国皇帝朱元璋曾盛赞《道德经》为"万物之根"，认为这部道家经典蕴含着治国安邦的至理。这位出身寒微的帝王在建立大明王朝后，深刻体会到老子"无为而治"的智慧，将其奉为治国圭臬。朱元璋不仅亲自批注《道德经》，更命翰林学士编修《御注道德经》，将其中"治大国若烹小鲜"等思想运用于轻徭薄赋的惠民政策中';
+// 按15个字分割文本为数组
+const text2Arr = [];
+for (let i = 0; i < text2.length; i += 2) {
+  text2Arr.push(text2.slice(i, i + 2));
+}
+const sleep = async (ms = 1000) => {
+  return new Promise((resolve) => {
+    setTimeout(() => {
+      resolve(true);
+    }, ms);
+  });
+};
+const mockSendText = (ttsMax: TtsMix) => {
+  return new Promise(async (resolve, reject) => {
+    for (let i = 0; i < text2Arr.length; i++) {
+      const text = text2Arr[i];
+      console.log('开始', i, text);
+      ttsMax.emitter.emit('text', text);
+      await sleep(10);
+      // console.log('完成', i, text);
+    }
+    ttsMax.emitter.emit('textEnd');
+  });
+};
 // tsx src/tts/provider/volcengine/test/mix.ts
 const main = async () => {
-  await runDemo(appId, token, speaker, text, outputPath);
+  try {
+    console.log('开始', appId, token);
+    // await runDemo(appId, token, speaker, text, outputPath);
+    const ttsMax = new TtsMix(appId, token);
+    setTimeout(() => {
+      mockSendText(ttsMax).then(() => {
         console.log('完成');
+      });
+    }, 10000);
+    // await ttsMax.getVoiceDemo(speaker, text, outputPath);
+
+    await ttsMax.getVoiceDemo(speaker, '', outputPath, false);
+    console.log('完成');
+  } catch (err) {
+    console.log(err);
+  }
 };
 
 main();

src/tts/provider/volcengine/tts-mix.ts

@@ -286,7 +286,7 @@ async function startSession(websocket: WebSocket, speaker: string, sessionId: st
 
 async function sendText(ws: WebSocket, speaker: string, text: string, sessionId: string): Promise<void> {
   const header = new Header(PROTOCOL_VERSION, DEFAULT_HEADER_SIZE, FULL_CLIENT_REQUEST, MsgTypeFlagWithEvent, JSON_TYPE).asBytes();
-
+  console.log('sendText=========', text);
   const optional = new Optional(EVENT_TaskRequest, sessionId).asBytes();
   const payload = getPayloadBytes('1234', EVENT_TaskRequest, text, speaker);
   return await sendEvent(ws, header, optional, payload);
@@ -308,7 +308,20 @@ async function finishConnection(ws: WebSocket): Promise<void> {
   return await sendEvent(ws, header, optional, payload);
 }
 
-export async function runDemo(appId: string, token: string, speaker: string, text: string, outputPath: string, emitter?: EventEmitter): Promise<void> {
+type RunOptions = {
+  id?: string;
+  autoEnd?: boolean;
+};
+export async function runDemo(
+  appId: string,
+  token: string,
+  speaker: string,
+  text: string,
+  outputPath: string,
+  emitter?: EventEmitter,
+  opts: RunOptions = {},
+): Promise<void> {
+  const autoEnd = opts.autoEnd ?? true;
   return new Promise((resolve, reject) => {
     const wsHeader = {
       'X-Api-App-Key': appId,
@@ -320,6 +333,7 @@ export async function runDemo(appId: string, token: string, speaker: string, tex
     const url = 'wss://openspeech.bytedance.com/api/v3/tts/bidirection';
     const ws = new WebSocket(url, { headers: wsHeader });
     const filename = outputPath.split('/').pop() || '';
+    // 开始连接
     let isBegin = true;
     const writeFileEmitter = (data: Buffer) => {
       const value: TTSWriteType = {
@@ -348,7 +362,19 @@ export async function runDemo(appId: string, token: string, speaker: string, tex
         let fileHandle: fs.FileHandle | null = null;
         let sessionId: string = '';
         let isFirstResponse = true;
-
+        let cacheText = '';
+        const emitOk = (id: string, code = 200) => {
+          emitter.emit(id, { code, msg: 'ok' });
+        };
+        emitter.on('text', async ({ text, id }) => {
+          await sendText(ws, speaker, text, sessionId);
+          emitOk(id);
+        });
+        emitter.on('textEnd', async ({ id }) => {
+          console.log('text end');
+          await finishSession(ws, sessionId);
+          emitOk(id);
+        });
         ws.on('message', async (data) => {
           try {
             const res = parserResponse(data as Buffer);
@@ -364,8 +390,10 @@ export async function runDemo(appId: string, token: string, speaker: string, tex
             }
             if (res.optional.event === EVENT_SessionStarted && isFirstResponse) {
               isFirstResponse = false;
-              await sendText(ws, speaker, text, sessionId);
+              console.log('start session', sessionId, autoEnd);
-              await finishSession(ws, sessionId);
+              emitter.emit('isConnect', sessionId);
+              text && (await sendText(ws, speaker, text, sessionId));
+              autoEnd && (await finishSession(ws, sessionId));
               fileHandle = await fs.open(outputPath, 'w');
             } else if (!isFirstResponse) {
               if (res.optional.event === EVENT_TTSResponse && res.header.messageType === AUDIO_ONLY_RESPONSE && res.payload && fileHandle) {
@@ -413,11 +441,14 @@ export class TtsMix {
   appId: string;
   token: string;
   emitter: EventEmitter;
-
+  isStart = false;
   constructor(appId: string, token: string) {
     this.appId = appId;
     this.token = token;
     this.emitter = new EventEmitter();
+    this.emitter.on('isConnect', () => {
+      this.isStart = true;
+    });
   }
   /**
    * 获取语音
@@ -426,8 +457,32 @@ export class TtsMix {
    * @param outputPath 输出路径
    * @returns
    */
-  async getVoiceDemo(speaker: string, text: string, outputPath: string): Promise<void> {
+  async getVoiceDemo(speaker: string, text: string, outputPath: string, autoEnd = true): Promise<void> {
-    return runDemo(this.appId, this.token, speaker, text, outputPath, this.emitter);
+    const id = nanoid();
+    const listenId = 'text' + id;
+    return runDemo(this.appId, this.token, speaker, text, outputPath, this.emitter, { autoEnd, id: listenId });
+  }
+  async isConnect() {
+    if (this.isStart) {
+      return Promise.resolve(true);
+    }
+    return new Promise((resolve) => {
+      this.emitter.once('isConnect', resolve);
+    });
+  }
+  async sendText(text: string): Promise<{ code?: number; msg?: string }> {
+    const id = nanoid();
+    return new Promise((resolve) => {
+      this.emitter.once(id, resolve);
+      this.emitter.emit('text', { text, id });
+    });
+  }
+  async sendTextEnd() {
+    const id = nanoid();
+    return new Promise((resolve) => {
+      this.emitter.once(id, resolve);
+      this.emitter.emit('textEnd', { id });
+    });
   }
   /**
    * 写入文件的时候同步的流监听