upd30

app.py

@@ -3,16 +3,14 @@ from fastapi.responses import StreamingResponse, JSONResponse
 import outetts
 import io
 import json
-import asyncio
-import os
-
+import base64
 # 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,
     )
@@ -27,40 +25,46 @@ 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()
-            # Schedule processing without awaiting
-            task = asyncio.create_task(process_chunk(data, websocket))
-            tasks.add(task)
-            task.add_done_callback(lambda t: tasks.discard(t))
+            # Status: Warming up
+            await websocket.send_text(json.dumps({"generation_status": "Warming up TTS model"}))
+            output = interface.generate(
+                config=outetts.GenerationConfig(
+                    text=data,
+                    generation_type=outetts.GenerationType.CHUNKED,
+                    speaker=speaker,
+                    sampler_config=outetts.SamplerConfig(
+                        temperature=0.4
+                    ),
+                )
+            )
+            # Status: Generating linguistic features
+            await websocket.send_text(json.dumps({"generation_status": "Generating linguistic features"}))
+            # Stream audio chunks
+            for chunk in output.stream(chunk_size=4096):
+                audio_b64 = base64.b64encode(chunk).decode("ascii")
+                await websocket.send_text(json.dumps({
+                    "data": {
+                        "audio_bytes": audio_b64,
+                        "duration": None,
+                        "request_finished": False
+                    }
+                }))
+            # Final event
+            await websocket.send_text(json.dumps({
+                "data": {
+                    "audio_bytes": "",
+                    "duration": None,
+                    "request_finished": True
+                }
+            }))
     except WebSocketDisconnect:
-        # Cancel all pending tasks
-        for task in tasks:
-            task.cancel()
+        pass
+    except Exception as e:
+        await websocket.send_text(json.dumps({"error": str(e)}))