Update gradio_app.py

gradio_app.py

@@ -126,7 +126,8 @@ def separate_speakers_video(video_path):
     output_files = []
     for i in range(ests_speech.shape[0]):
         separated_audio_path = os.path.join(output_dir, f"speaker_{i+1}.wav")
-        torchaudio.save(separated_audio_path, ests_speech[i].unsqueeze(0).cpu(), TARGET_SR)
+        mono_audio = ests_speech[i].cpu().unsqueeze(0)  # Shape: [1, time]
+        torchaudio.save(separated_audio_path, mono_audio.contiguous(), TARGET_SR, format="wav", encoding="PCM_S")  # safest combo
 
         # Attach audio back to video
         out_video_path = os.path.join(output_dir, f"speaker_{i+1}.mp4")