Update voice_processing.py

voice_processing.py

@@ -92,20 +92,22 @@ def process_audio(model, audio_file, logger, index_rate=0, use_uploaded_voice=Tr
     try:
         logger.info("Starting audio processing")
         
-        if model is None:
-            logger.error("No model provided for processing")
-            return None
+        # Load audio using librosa directly (matching original working code)
+        audio, sr = librosa.load(audio_file, sr=16000, mono=True)
+        logger.info(f"Loaded audio: sr={sr}Hz, shape={audio.shape}")
 
-        # Load and process audio
+        # Get model data using existing function
         tgt_sr, net_g, vc, version, index_file, if_f0 = model_data(model_name)
-        if f0_method == "rmvpe":
-            vc.model_rmvpe = rmvpe_model
+        
+        # Set RMVPE
+        vc.model_rmvpe = rmvpe_model
 
+        # Process using the VC pipeline that we know works
         times = [0, 0, 0]
         audio_opt = vc.pipeline(
-            hubert_model,
-            net_g,
-            0,  # sid
+            hubert_model,  # Use global hubert model
+            net_g,  # Use the generator from model_data
+            0,  # speaker id
             audio,
             audio_file,
             times,
@@ -123,6 +125,9 @@ def process_audio(model, audio_file, logger, index_rate=0, use_uploaded_voice=Tr
             f0_file=None
         )
 
+        if tgt_sr != 0 and tgt_sr >= 16000:
+            tgt_sr = resample_sr
+
         info = f"Success. Time: npy: {times[0]}s, f0: {times[1]}s, infer: {times[2]}s"
         logger.info(info)
         return (info, None, (tgt_sr, audio_opt))