Updated Gradio App

app.py

@@ -28,17 +28,13 @@ def process_audio(audio):
     if audio_data is None or len(audio_data) == 0:
         return "Audio data is empty. Please try again with a valid audio file."
     
-    # Ensure audio_data is a numpy array.
-    audio_data = np.asarray(audio_data)
+    # Force conversion of audio data to a floating-point numpy array.
+    audio_data = np.array(audio_data, dtype=np.float32)
     
-    # If audio data is multi-dimensional, squeeze to 1D.
+    # If the audio data is multi-dimensional, squeeze it to 1D.
     if audio_data.ndim > 1:
         audio_data = np.squeeze(audio_data)
     
-    # Convert audio data to floating-point if it's not already.
-    if not np.issubdtype(audio_data.dtype, np.floating):
-        audio_data = audio_data.astype(np.float32)
-    
     # Resample to 16000 Hz if necessary.
     if sample_rate != 16000:
         try:
@@ -69,12 +65,12 @@ def process_audio(audio):
 # Create the Gradio interface.
 iface = gr.Interface(
     fn=process_audio,
-    inputs=gr.Audio(type="numpy"),  # Using file upload for audio input.
+    inputs=gr.Audio(type="numpy"),  # Use file upload for audio input.
     outputs="text",
     title="Sarvam AI Shuka Voice Demo",
     description="Upload an audio file and get a response using Sarvam AI's Shuka model."
 )
 
 if __name__ == "__main__":
-    # Set share=True to create a public link and use a non-default port.
-    iface.launch(share=True, server_port=7861)
+    # Launch the app with share=True to create a public link.
+    iface.launch(share=True)