阿里云一句话识别

package.json

@@ -49,6 +49,7 @@
     "vosk": "^0.3.39"
   },
   "devDependencies": {
+    "@alicloud/pop-core": "^1.8.0",
     "@kevisual/logger": "^0.0.3",
     "@kevisual/types": "^0.0.6",
     "@kevisual/use-config": "^1.0.10",

src/asr/provider/aliyun/aliyun-asr-server.ts

@@ -0,0 +1,131 @@
+type AliAsrServerOptions = {
+  baseUrl?: string;
+  appkey: string;
+  token: string;
+  format?: string;
+  sampleRate?: string;
+  enablePunctuationPrediction?: boolean;
+  enableInverseTextNormalization?: boolean;
+  enableVoiceDetection?: boolean;
+};
+export class AliAsrServer {
+  private baseUrl: string;
+  private appkey: string;
+  private token: string;
+  private format: string;
+  private sampleRate: string;
+  private enablePunctuationPrediction: boolean;
+  private enableInverseTextNormalization: boolean;
+  private enableVoiceDetection: boolean;
+
+  constructor(opts?: AliAsrServerOptions) {
+    const {
+      baseUrl = 'https://nls-gateway-cn-shanghai.aliyuncs.com/stream/v1/asr',
+      appkey = '',
+      token = '',
+      format,
+      sampleRate,
+      enablePunctuationPrediction = true,
+      enableInverseTextNormalization = true,
+      enableVoiceDetection = false,
+    } = opts || {};
+    this.baseUrl = baseUrl;
+    this.appkey = appkey;
+    this.token = token;
+    this.format = format;
+    this.sampleRate = sampleRate;
+    this.enablePunctuationPrediction = enablePunctuationPrediction;
+    this.enableInverseTextNormalization = enableInverseTextNormalization;
+    this.enableVoiceDetection = enableVoiceDetection;
+  }
+
+  buildRequestUrl(): string {
+    const params = new URLSearchParams();
+    params.append('appkey', this.appkey);
+    this.format && params.append('format', this.format);
+    this.sampleRate && params.append('sample_rate', this.sampleRate);
+
+    if (this.enablePunctuationPrediction) {
+      params.append('enable_punctuation_prediction', 'true');
+    }
+
+    if (this.enableInverseTextNormalization) {
+      params.append('enable_inverse_text_normalization', 'true');
+    }
+
+    if (this.enableVoiceDetection) {
+      params.append('enable_voice_detection', 'true');
+    }
+
+    return `${this.baseUrl}?${params.toString()}`;
+  }
+
+  async processAudio(audioContent: Buffer): Promise<any> {
+    try {
+      // 设置请求头
+      const headers = {
+        'X-NLS-Token': this.token,
+        'Content-Type': 'application/octet-stream',
+      };
+
+      // 构建请求URL
+      const requestUrl = this.buildRequestUrl();
+
+      // 发送请求
+      const response = await fetch(requestUrl, {
+        method: 'POST',
+        headers,
+        body: audioContent,
+      });
+
+      // 处理响应
+      if (!response.ok) {
+        console.log(`The audio file recognized failed, http code: ${response.status}`);
+        const v = await response.text();
+        console.log('The audio file recognized response:', v);
+        return null;
+      }
+      // 解析响应体
+      // console.log('The audio file recognized response:', v);
+      const body = await response.json();
+
+      if (body.status === 20000000) {
+        console.log('The audio file recognized result:');
+        console.log(body);
+        console.log('result: ' + body.result);
+        console.log('The audio file recognized succeed!');
+        return body;
+      } else {
+        console.log('The audio file recognized failed!');
+        console.log(body);
+        return null;
+      }
+    } catch (error) {
+      if (error.code === 'ENOENT') {
+        console.log('The audio file does not exist!');
+      } else {
+        console.log('Error during audio processing:', error);
+      }
+      return null;
+    }
+  }
+}
+
+// // 使用示例
+// async function main() {
+//   const asrServer = new AliAsrServer({
+//     appkey: '填入appkey',
+//     token: '填入服务鉴权Token',
+//     format: 'pcm',
+//     sampleRate: '16000',
+//     enablePunctuationPrediction: true,
+//     enableInverseTextNormalization: true,
+//     enableVoiceDetection: false,
+//   });
+
+//   const audioFile = '/path/to/nls-sample-16k.wav';
+//   await asrServer.processAudio(audioFile);
+// }
+
+// // 执行主函数
+// main().catch(console.error);

src/asr/provider/aliyun/base.ts

@@ -0,0 +1,42 @@
+import RPCClient from '@alicloud/pop-core';
+
+interface TokenResponse {
+  Token: {
+    Id: string;
+    ExpireTime: number;
+  };
+}
+type AliCommonOptions = {
+  accessKeyId: string;
+  accessKeySecret: string;
+};
+export class AliCommon {
+  private accessKeyId: string;
+  private accessKeySecret: string;
+  private endpoint: string;
+  private apiVersion: string;
+  token = '';
+  expireTime = 0;
+  constructor(opts?: AliCommonOptions) {
+    this.accessKeyId = opts?.accessKeyId || process.env.ALIYUN_AK_ID || '';
+    this.accessKeySecret = opts?.accessKeySecret || process.env.ALIYUN_AK_SECRET || '';
+    this.endpoint = 'http://nls-meta.cn-shanghai.aliyuncs.com';
+    this.apiVersion = '2019-02-28';
+  }
+  async getToken() {
+    if (this.token && this.expireTime > Date.now()) {
+      return this.token;
+    }
+    const client = new RPCClient({
+      accessKeyId: this.accessKeyId,
+      accessKeySecret: this.accessKeySecret,
+      endpoint: this.endpoint,
+      apiVersion: this.apiVersion,
+    });
+
+    const result = await client.request<TokenResponse>('CreateToken', {});
+    this.token = result.Token.Id;
+    this.expireTime = result.Token.ExpireTime * 1000;
+    return result.Token.Id;
+  }
+}

src/asr/provider/aliyun/test/get-text.ts

@@ -0,0 +1,25 @@
+import { AliAsrServer } from '../aliyun-asr-server.ts';
+import fs from 'fs/promises';
+import path from 'path';
+// const videoTestPath = path.join(process.cwd(), 'videos/asr_example.wav');
+// const videoTestPath = path.join(process.cwd(), 'videos/asr_example2.wav');
+const videoTestPath = path.join(process.cwd(), 'videos/tts_mix.mp3');
+// const videoTestPath = path.join(process.cwd(), 'videos/my_speech_text.wav');
+const name = 'output-1746007775571.mp3';
+const videoTestPath2 = path.join(process.cwd(), 'build', name);
+
+// 使用示例
+async function main() {
+  const asrServer = new AliAsrServer({
+    appkey: process.env.ALI_ASR_APP_KEY,
+    token: process.env.ALI_ASR_TOKEN,
+    format: 'mp3',
+    // format: 'wav',
+  });
+
+  const audioContent = await fs.readFile(videoTestPath);
+  await asrServer.processAudio(audioContent);
+}
+
+// 执行主函数
+main().catch(console.error);

src/asr/provider/aliyun/test/get-token.ts

@@ -0,0 +1,10 @@
+import dotenv from 'dotenv';
+dotenv.config();
+import { AliCommon } from '../base.ts';
+
+const aliCommon = new AliCommon({
+  accessKeyId: process.env.ALIYUN_AK_ID,
+  accessKeySecret: process.env.ALIYUN_AK_SECRET,
+});
+
+aliCommon.getToken().then(console.log);

src/asr/provider/funasr/test/get-text.ts

@@ -8,8 +8,9 @@ import fs from 'fs';
 const videoTestPath = path.join(process.cwd(), 'videos/tts_mix.mp3');
 // const videoTestPath = path.join(process.cwd(), 'videos/my_speech_text.wav');
 const name = 'output-1746007775571.mp3';
-const url = 'wss://funasr.xiongxiao.me';
 const videoTestPath2 = path.join(process.cwd(), 'build', name);
+
+const url = 'wss://funasr.xiongxiao.me';
 // const ws = new VideoWS({
 //   // url: 'wss://192.168.31.220:10095',
 //   url: 'wss://funasr.xiongxiao.me',