multi channel input, warn user and convert

app.py

@@ -182,6 +182,9 @@ def prepare_prompt(text: str, speaker: int, audio_path: str) -> Segment:
 
 def load_prompt_audio(audio_path: str) -> torch.Tensor:
     audio_tensor, sample_rate = torchaudio.load(audio_path)
+    if audio_tensor.shape[0] != 1:
+        gr.Warning("Warning: Audio prompt is multi-channel, converting to mono.", duration=15)
+        audio_tensor = audio_tensor.mean(dim=0)
     audio_tensor = audio_tensor.squeeze(0)
     if sample_rate != generator.sample_rate:
         audio_tensor = torchaudio.functional.resample(