Update app.py

app.py

@@ -104,10 +104,9 @@ def transcribe_audio(audio_file, chunk_length_s=30):
         if chunk_waveform.shape[0] > 1:
             chunk_waveform = torch.mean(chunk_waveform, dim=0, keepdim=True)
 
-# Process the chunk with the tokenizer
-inputs = processor(chunk_waveform.squeeze(0).numpy(), sampling_rate=sample_rate, return_tensors="pt")
-
-input_features = inputs.input_features
+        # Process the chunk with the tokenizer
+        inputs = processor(chunk_waveform.squeeze(0).numpy(), sampling_rate=sample_rate, return_tensors="pt")
+        input_features = inputs.input_features
 
 # Create attention mask
 attention_mask = torch.ones(inputs.input_features.shape[:2], dtype=torch.long, device=device)