Update custom_interface_app.py

custom_interface_app.py

@@ -93,8 +93,8 @@ class ASR(Pretrained):
         sr = 16000
         max_segment_length = 30
 
-        # 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)