Update voice_analysis.py

voice_analysis.py

@@ -33,6 +33,7 @@ def get_speaker_embeddings(audio_path, diarization, model_name="pyannote/embeddi
     model.eval()  # Set the model to evaluation mode
     
     waveform, sample_rate = torchaudio.load(audio_path)
+    print(f"Sample rate: {sample_rate}")
     embeddings = []
 
     for turn, _, speaker in diarization.itertracks(yield_label=True):
@@ -40,6 +41,8 @@ def get_speaker_embeddings(audio_path, diarization, model_name="pyannote/embeddi
         end_frame = int(turn.end * sample_rate)
         
         segment = waveform[:, start_frame:end_frame]
+        print(f"Segment shape before processing: {segment.shape}")
+        
         if segment.shape[1] == 0:
             continue
 
@@ -54,6 +57,8 @@ def get_speaker_embeddings(audio_path, diarization, model_name="pyannote/embeddi
 
         # Reshape the segment to match the model's expected input
         segment = segment.unsqueeze(0)  # Add batch dimension
+        
+        print(f"Segment shape after processing: {segment.shape}")
 
         with torch.no_grad():
             embedding = model(segment)  # Pass the tensor directly, not a dictionary