tts for cosyvoice and funasr and aliyun

package.json

@@ -36,6 +36,7 @@
     "access": "public"
   },
   "dependencies": {
+    "@gradio/client": "^1.14.2",
     "@kevisual/router": "0.0.10",
     "@kevisual/use-config": "^1.0.10",
     "@kevisual/video": "^0.0.1",

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

@@ -3,8 +3,8 @@ 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 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);
 

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

@@ -5,8 +5,8 @@ import fs from 'fs';
 
 // 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 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);
 

src/tts/provider/cosyvoice/test/tts.ts

@@ -0,0 +1,28 @@
+import { Client } from '@gradio/client';
+import path from 'node:path';
+import fs from 'node:fs';
+
+// 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);
+const exampleAudio = fs.readFileSync(videoTestPath);
+// const exampleAudio = await response_0.blob();
+
+const client = await Client.connect('http://192.168.31.220:50000/');
+const result = await client.predict('/generate_audio', {
+  tts_text: '梦里的一天结束了一个新的轮m,回梦里的一天结束了一个新的轮回梦里的一,,天结束了一个新的轮回,梦里的一天结束了一个新的轮回',
+  mode_checkbox_group: '3s极速复刻',
+  sft_dropdown: '',
+  prompt_text: '在一无所知中，梦里的一天结束了一个新的轮回，便会开始。',
+  prompt_wav_upload: exampleAudio,
+  prompt_wav_record: null,
+  instruct_text: '',
+  seed: 3,
+  stream: false,
+  speed: 1,
+});
+
+console.log(result.data);

src/tts/provider/cosyvoice/tts.ts

@@ -0,0 +1,53 @@
+import { Client } from '@gradio/client';
+
+type CosyVoiceTTSOptions = {
+  url: string;
+};
+type AudioOptions = {
+  tts_text: string;
+  mode_checkbox_group: string;
+  sft_dropdown: string;
+  prompt_text: string;
+  prompt_wav_upload?: any;
+  prompt_wav_record: any | null;
+  instruct_text: string;
+  seed: number;
+  stream: boolean;
+  speed: number;
+};
+
+export class CosyVoiceTTS {
+  private client: Client;
+  private url: string;
+  isInit = false;
+  constructor(opts?: CosyVoiceTTSOptions) {
+    this.url = opts?.url || 'http://localhost:50000/';
+  }
+  async init() {
+    const url = this.url;
+    const client = await Client.connect(url);
+    this.client = client;
+    this.isInit = true;
+    return true;
+  }
+  generateAudio = async (opts?: Partial<AudioOptions>) => {
+    if (!this.isInit) {
+      await this.init();
+    }
+    const data: AudioOptions = {
+      tts_text: '梦里的一天结束了一个新的轮m,回梦里的一天结束了一个新的轮回梦里的一,,天结束了一个新的轮回,梦里的一天结束了一个新的轮回',
+      mode_checkbox_group: '3s极速复刻',
+      sft_dropdown: '',
+      prompt_text: '在一无所知中，梦里的一天结束了一个新的轮回，便会开始。',
+      // prompt_wav_upload: exampleAudio,
+      prompt_wav_record: null,
+      instruct_text: '',
+      seed: 3,
+      stream: false,
+      speed: 1,
+      ...opts,
+    };
+    const result = await this.client.predict('/generate_audio', data);
+    return result;
+  };
+}