Update app.py

app.py

@@ -101,6 +101,9 @@ def infer_img2img(prompt, audio_path, progress=gr.Progress(track_tqdm=True)):
     # Ensure correct shape
     concat_audio = concat_audio.flatten()  # Converts (1, N) → (N,)
 
+    # Normalize the audio to prevent clipping or excessive loudness
+    concat_audio = concat_audio / np.max(np.abs(concat_audio))  # Scale between -1 and 1
+
     # Save as WAV
     sf.write("output.wav", concat_audio, 16000)