Update app.py

app.py

@@ -3,13 +3,16 @@ from fastapi.responses import StreamingResponse, JSONResponse
 import outetts
 import io
 import json
+import asyncio
+import os
+
 # Initialize the interface
 interface = outetts.Interface(
     config=outetts.ModelConfig.auto_config(
         model=outetts.Models.VERSION_1_0_SIZE_1B,
         # For llama.cpp backend
-        #backend=outetts.Backend.LLAMACPP,
-        #quantization=outetts.LlamaCppQuantization.FP16
+        # backend=outetts.Backend.LLAMACPP,
+        # quantization=outetts.LlamaCppQuantization.FP16
         # For transformers backend
         backend=outetts.Backend.HF,
     )
@@ -24,32 +27,40 @@ app = FastAPI()
 def greet_json():
     return {"Hello": "World!"}
 
+async def process_chunk(text_chunk: str, websocket: WebSocket):
+    try:
+        output = interface.generate(
+            config=outetts.GenerationConfig(
+                text=text_chunk,
+                generation_type=outetts.GenerationType.CHUNKED,
+                speaker=speaker,
+                sampler_config=outetts.SamplerConfig(
+                    temperature=0.4
+                ),
+            )
+        )
+        # Save audio to buffer
+        audio_buffer = io.BytesIO()
+        output.save(audio_buffer)
+        audio_buffer.seek(0)
+        audio_bytes = audio_buffer.read()
+        # Send audio bytes back
+        await websocket.send_bytes(audio_bytes)
+    except Exception as e:
+        await websocket.send_text(json.dumps({"error": str(e)}))
+
 @app.websocket("/ws/tts")
 async def websocket_tts(websocket: WebSocket):
     await websocket.accept()
+    tasks: set[asyncio.Task] = set()
     try:
         while True:
-            # Empfange Text-Chunk vom Client
             data = await websocket.receive_text()
-            # Generiere Audio aus Text
-            output = interface.generate(
-                config=outetts.GenerationConfig(
-                    text=data,
-                    generation_type=outetts.GenerationType.CHUNKED,
-                    speaker=speaker,
-                    sampler_config=outetts.SamplerConfig(
-                        temperature=0.4
-                    ),
-                )
-            )
-             # Speichere Audio temporär als Datei
-            temp_path = "temp.wav"
-            output.save(temp_path)
-            with open(temp_path, "rb") as f:
-                audio_bytes = f.read()
-            import os
-            os.remove(temp_path)
-            # Sende Audiodaten als Bytes zurück
-            await websocket.send_bytes(audio_bytes)
+            # Schedule processing without awaiting
+            task = asyncio.create_task(process_chunk(data, websocket))
+            tasks.add(task)
+            task.add_done_callback(lambda t: tasks.discard(t))
     except WebSocketDisconnect:
-        pass
+        # Cancel all pending tasks
+        for task in tasks:
+            task.cancel()