Upload 4 files

tts-server.py

@@ -244,7 +244,7 @@ class DoubaoTTS:
         except Exception as e:
             logger.error(f"Failed to save models.json: {e}")
 
-    async def stream_audio(self, text: str, voice: str, speed: float = 1.0, pitch: float = 1.0) -> AsyncGenerator[bytes, None]:
+    async def stream_audio(self, text: str, voice: str, format: str = "aac", speed: float = 1.0, pitch: float = 1.0) -> AsyncGenerator[bytes, None]:
         """Connect to WebSocket and yield audio chunks with retry logic."""
         
         # Map OpenAI speed (0.25 - 4.0) to Doubao rate (-100 to 100)
@@ -262,7 +262,7 @@ class DoubaoTTS:
                 return
 
             params = self._get_common_params()
-            ws_url = f"{self.ws_url}?format=aac&speaker={voice}&speech_rate={doubao_rate}&pitch={doubao_pitch}{params}"
+            ws_url = f"{self.ws_url}?format={format}&speaker={voice}&speech_rate={doubao_rate}&pitch={doubao_pitch}{params}"
             
             headers = {
                 "Cookie": cookie,
@@ -359,8 +359,15 @@ async def create_speech(req: OpenAIRequest, token: str = Depends(verify_token)):
     if req.response_format == "mp3":
         media_type = "audio/mpeg"
     
+    # Determine format to request from Doubao
+    target_format = "aac"
+    if req.response_format == "mp3":
+        target_format = "mp3"
+    elif req.response_format == "pcm":
+        target_format = "pcm"
+
     return StreamingResponse(
-        engine.stream_audio(req.input, req.voice, req.speed, req.pitch),
+        engine.stream_audio(req.input, req.voice, target_format, req.speed, req.pitch),
         media_type=media_type
     )