fix: ensure MP3 to WAV conversion creates file

- Add proper error handling around pydub conversion
- Use absolute paths instead of relative paths
- Ensure output directory exists before saving
- Verify file was actually created before returning
- Add fallback to MP3 if conversion fails
- Update tts.py to handle both .wav and .mp3 extensions
- Add detailed logging for debugging

app/api/tts.py

@@ -121,16 +121,28 @@ class TTS(Resource):
             seed_vc_model = current_app.config['SEED_VC_MODEL']
             seed_vc_repo = current_app.config['SEED_VC_REPO_PATH']
             vc_service = SeedVCService(model_name=seed_vc_model, repo_path=seed_vc_repo)
-            vc_service.convert_voice(tts_output, reference_path, final_output)
+            actual_output = vc_service.convert_voice(tts_output, reference_path, final_output)
             
-            logger.info(f"TTS+VC complete: {final_output}")
+            # Verify the output file exists
+            if not os.path.exists(actual_output):
+                raise FileNotFoundError(f"Output file was not created: {actual_output}")
+            
+            # Determine mimetype and download name based on actual output
+            if actual_output.endswith('.mp3'):
+                mimetype = 'audio/mpeg'
+                download_name = f"tts_{voice_id}.mp3"
+            else:
+                mimetype = 'audio/wav'
+                download_name = f"tts_{voice_id}.wav"
+            
+            logger.info(f"TTS+VC complete: {actual_output} ({os.path.getsize(actual_output)} bytes)")
             
             # Return audio file
             return send_file(
-                final_output,
-                mimetype='audio/wav',
+                actual_output,
+                mimetype=mimetype,
                 as_attachment=True,
-                download_name=f"tts_{voice_id}.wav"
+                download_name=download_name
             )
             
         except Exception as e:

app/services/seed_vc_service.py

@@ -92,12 +92,7 @@ class SeedVCService:
         if self.model is None:
             # Fallback: convert MP3 to WAV
             logger.warning("Seed-VC not available, using passthrough mode")
-            from pydub import AudioSegment
-            # Load MP3 and convert to WAV
-            audio = AudioSegment.from_mp3(source_audio_path)
-            audio.export(output_path, format="wav")
-            logger.info(f"MP3 to WAV conversion complete: {output_path}")
-            return output_path
+            return self._passthrough_convert(source_audio_path, output_path)
         
         try:
             # Load audio files
@@ -122,11 +117,76 @@ class SeedVCService:
             logger.error(f"Voice conversion failed: {e}")
             # Fallback: convert MP3 to WAV
             logger.warning("Fallback to MP3-to-WAV conversion")
+            return self._passthrough_convert(source_audio_path, output_path)
+    
+    def _passthrough_convert(self, source_audio_path: str, output_path: str) -> str:
+        """
+        Convert MP3 to WAV as passthrough when Seed-VC is not available.
+        
+        Args:
+            source_audio_path: Path to source MP3 audio
+            output_path: Path to save converted WAV audio
+        
+        Returns:
+            Path to the output file (may differ from output_path if MP3 is used)
+        """
+        # Use absolute paths
+        source_audio_path = os.path.abspath(source_audio_path)
+        output_path = os.path.abspath(output_path)
+        
+        logger.info(f"Passthrough convert: {source_audio_path} -> {output_path}")
+        
+        # Verify source exists
+        if not os.path.exists(source_audio_path):
+            raise FileNotFoundError(f"Source audio file not found: {source_audio_path}")
+        
+        # Ensure output directory exists
+        output_dir = os.path.dirname(output_path)
+        if output_dir:
+            os.makedirs(output_dir, exist_ok=True)
+            logger.debug(f"Ensured output directory exists: {output_dir}")
+        
+        try:
             from pydub import AudioSegment
+            
+            # Load MP3
+            logger.debug(f"Loading MP3 from: {source_audio_path}")
             audio = AudioSegment.from_mp3(source_audio_path)
+            logger.debug(f"Audio loaded: {len(audio)}ms, {audio.channels} channels")
+            
+            # Export as WAV
+            logger.debug(f"Exporting WAV to: {output_path}")
             audio.export(output_path, format="wav")
-            logger.info(f"MP3 to WAV conversion complete: {output_path}")
+            
+            # Verify file was created
+            if not os.path.exists(output_path):
+                raise RuntimeError(f"WAV file was not created at {output_path}")
+            
+            file_size = os.path.getsize(output_path)
+            if file_size == 0:
+                raise RuntimeError(f"WAV file is empty: {output_path}")
+            
+            logger.info(f"MP3 to WAV conversion complete: {output_path} ({file_size} bytes)")
             return output_path
+            
+        except Exception as e:
+            logger.error(f"MP3 to WAV conversion failed: {e}")
+            
+            # Fallback: return MP3 file directly (rename extension)
+            mp3_output = output_path.replace('.wav', '.mp3')
+            if source_audio_path != mp3_output:
+                import shutil
+                logger.warning(f"Using MP3 fallback: copying to {mp3_output}")
+                shutil.copy2(source_audio_path, mp3_output)
+                
+                if not os.path.exists(mp3_output):
+                    raise RuntimeError(f"Failed to create fallback MP3 file: {mp3_output}")
+                
+                logger.info(f"Fallback MP3 created: {mp3_output}")
+                return mp3_output
+            else:
+                # Source is already the target
+                return source_audio_path
     
     def preprocess_audio(self, input_path: str, output_path: str, target_sr: int = 22050) -> str:
         """