added mic input

app.py

@@ -28,10 +28,19 @@ _ = net_g.eval()
 _ = utils.load_checkpoint("logs/ow2/G_195000.pth", net_g, None)
 
 
-def infer(md, audio, speaker_id, pitch_shift, length_scale, noise_scale=.667, noise_scale_w=0.8):
-    fname = audio
-    source, sr = torchaudio.load(fname)
+def infer(md, mic_audio, audio, speaker_id, pitch_shift, length_scale, noise_scale=.667, noise_scale_w=0.8):
     
+    source = None
+    sr = None
+    
+    if mic_audio:
+        sr, source  = mic_audio
+        source = torch.Tensor(source)
+        source = source.T
+
+    if audio:
+        source, sr = torchaudio.load(audio)
+
     source = torchaudio.functional.pitch_shift(source, sr, int(pitch_shift))#, n_fft=256)
     source = torchaudio.functional.resample(source, sr, 16000)
     source = torch.mean(source, dim=0).unsqueeze(0)
@@ -63,6 +72,7 @@ demo = gradio.Interface(
         
         ( * up to Kiriko and without Bastion. Please forgive. )
         """),
+        gradio.Audio(source="microphone", label="Record your own voice"),
         gradio.Audio(label="Input Audio", type="filepath"),
         gradio.Dropdown(label="Target Voice", choices=["Ana", "Ashe", "Baptiste", "Brigitte", "Cassidy", "Doomfist", "D.Va", "Echo", "Genji", "Hanzo", "Junker Queen", "Junkrat", "Kiriko", "Lúcio", "Mei", "Mercy", "Moira", "Orisa", "Pharah", "Reaper", "Reinhardt", "Roadhog", "Sigma", "Sojourn", "Soldier_ 76", "Sombra", "Symmetra", "Torbjörn", "Tracer", "Widowmaker", "Winston", "Zarya", "Zenyatta"], type="index", value="Ana"),
         gradio.Slider(label="Pitch Shift Input (+12 = up one octave, ⚠️ broken AF ⚠️)", minimum=-12.0, maximum=12.0, value=0, step=1),