Update custom_interface_app.py

custom_interface_app.py

@@ -85,12 +85,13 @@ class ASR(Pretrained):
 
         # waveform, sr = librosa.load(file, sr=sr)
         waveform, file_sr = torchaudio.load(file)
+        waveform = waveform.mean(dim=0, keepdim=True) # convert to mono
         # resample if not 16kHz
         if file_sr != sr:
             waveform = torchaudio.transforms.Resample(file_sr, sr)(waveform)
 
         # limit to 1 min
-        waveform = waveform[:, :60*sr]
+        # waveform = waveform[:, :60*sr]
         
         waveform = waveform.squeeze()
         audio_length = len(waveform) / sr