Update app.py

app.py

@@ -27,8 +27,8 @@ def transcribe(audio):
         resampler = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)
         audio_tensor = resampler(audio_tensor)
     
-    # Ensure the audio tensor is on the correct device
-    audio_tensor = audio_tensor.to(device)
+    # Convert back to a NumPy array
+    audio_np = audio_tensor.cpu().numpy()
     
     pipe = pipeline(
         "automatic-speech-recognition",
@@ -37,9 +37,8 @@ def transcribe(audio):
         device=device,
     )
     
-    prediction = pipe(audio_tensor)["text"]
+    prediction = pipe(audio_np)["text"]
     return prediction
-
     
 gradio_app = gr.Interface(
     fn=transcribe,