feat: realistic face avatar + voice input + ASR endpoint

backend/src/modules/tts/tts.service.ts

@@ -12,6 +12,18 @@ interface TtsResult {
   durationMs: number
 }
 
+const VALID_VOICES = new Set([
+  'zh-CN-XiaoxiaoNeural', 'zh-CN-XiaoyiNeural', 'zh-CN-YunjianNeural',
+  'zh-CN-YunxiNeural', 'zh-CN-YunxiaNeural', 'zh-CN-YunyangNeural',
+  'zh-CN-liaoning-XiaobeiNeural', 'zh-CN-shaanxi-XiaoniNeural',
+])
+
+function validateVoice(voice: string): void {
+  if (!VALID_VOICES.has(voice)) {
+    throw new Error(`不支持的语音: ${voice}`)
+  }
+}
+
 @Injectable()
 export class TtsService {
   private readonly logger = new Logger(TtsService.name)
@@ -23,6 +35,7 @@ export class TtsService {
   }
 
   async synthesize(text: string, voice: string = 'zh-CN-XiaoxiaoNeural'): Promise<TtsResult> {
+    validateVoice(voice)
     const hash = crypto.createHash('md5').update(text + voice).digest('hex')
     const filePath = path.join(CACHE_DIR, `${hash}.mp3`)