Update main.py

main.py

@@ -1,4 +1,4 @@
-from fastapi import FastAPI, HTTPException, BackgroundTasks
+from fastapi import FastAPI, HTTPException, BackgroundTasks, Request
 from fastapi.responses import FileResponse, JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
@@ -12,6 +12,7 @@ import logging
 import traceback
 from typing import Optional
 import torch
+import soundfile as sf
 
 # Configure logging
 logging.basicConfig(level=logging.INFO,
@@ -23,6 +24,7 @@ os.environ["OUTETTS_NO_PORTAUDIO"] = "1"
 
 # Import the TextToSpeech class from generate.py
 try:
+    sys.path.append(os.path.dirname(os.path.abspath(__file__)))
     from generate import TextToSpeech
     logger.info("Successfully imported TextToSpeech class from generate.py")
 except ImportError as e:
@@ -42,7 +44,7 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# YarnGPT TTS configuration - This can be adjusted based on model availability
+# YarnGPT TTS configuration
 MODEL_CONFIG = {
     "model_name_or_path": "yarngpt/yarn-tts-demo",
     "processor_name_or_path": "yarngpt/yarn-tts-demo"
@@ -144,23 +146,21 @@ async def text_to_speech(request: TTSRequest, background_tasks: BackgroundTasks)
         
         logger.info(f"Processing TTS request: '{request.text[:50]}...' with voice '{request.voice}' and language '{request.language}'")
         
+        # Create prompt from voice and language
+        # This adapts to the colab-style API even though we're using a different backend
+        accent = request.language if request.language in ["nigerian"] else "nigerian"
+        
         # Generate speech
         try:
-            # Map the language/voice to accent
-            # For simplicity we're using nigerian accent for all, but this could be enhanced
-            accent = "nigerian"
-            
-            # Generate audio data
             audio_data, sample_rate = yarngpt.tts(
                 text=request.text,
-                accent=accent, 
+                accent=accent,
                 save_path=output_path,
                 speed=request.speed,
                 get_array=True
             )
             
             # Convert audio to base64
-            import soundfile as sf
             sf.write(output_path, audio_data, sample_rate)
             with open(output_path, "rb") as audio_file:
                 audio_bytes = audio_file.read()
@@ -195,7 +195,7 @@ async def text_to_speech(request: TTSRequest, background_tasks: BackgroundTasks)
         traceback.print_exc()
         raise HTTPException(status_code=500, detail=f"Failed to generate speech: {str(e)}")
 
-# File serving endpoint (for compatibility with direct requests)
+# File serving endpoint (for backward compatibility)
 @app.get("/audio/{filename}")
 async def get_audio(filename: str):
     file_path = os.path.join(AUDIO_DIR, filename)
@@ -203,7 +203,6 @@ async def get_audio(filename: str):
         raise HTTPException(status_code=404, detail="Audio file not found")
     return FileResponse(file_path, media_type="audio/wav")
 
-# For local testing
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=True)