Update app.py

app.py

@@ -59,7 +59,19 @@ _ = utils.load_checkpoint("checkpoints/freevc-cvfr.pth", freevc_cvfr, None)
 print("Loading WavLM for content...")
 cmodel = WavLMModel.from_pretrained("microsoft/wavlm-large").to(device)
  
-def convert(model, src, tgt):
+def convert(model, src_mic,src_file, tgt):
+    """
+    helper function which checks where source come from
+    """
+    src = None
+    if src_mic:
+        src_= src_mic
+    elif src_file:
+        src = src_file
+    if not src:
+        logging.warning("source or target are not provided")
+        return
+ 
     with torch.no_grad():
         # tgt
         wav_tgt, _ = librosa.load(tgt, sr=hps.data.sampling_rate)
@@ -101,10 +113,10 @@ def convert(model, src, tgt):
     return out
     
 model = gr.Dropdown(choices=["FreeVC", "FreeVC-s", "FreeVC (24kHz)","FreeVC CVFR"], value="FreeVC",type="value", label="Model") 
-#audio1 = gr.inputs.Audio(label="Source Audio", type='filepath')
-audio1=gr.Audio(source="microphone", type="filepath")
+audio1_mic=gr.Audio(source="microphone", type="filepath", label='record your voice', optional=True)
+audio1_file = gr.inputs.Audio(label="Source Audio", type='filepath', label='or upload an audio file', optional=True)
 audio2 = gr.inputs.Audio(label="Reference Audio", type='filepath')
-inputs = [model, audio1, audio2]
+inputs = [model, audio1_mic, audio1_file, audio2]
 outputs =  gr.outputs.Audio(label="Output Audio", type='filepath')
 
 title = "FreeVC"