Update app.py

app.py

@@ -59,6 +59,15 @@ freevc_cvfr = SynthesizerTrn(
 _ = freevc_cvfr.eval()
 _ = utils.load_checkpoint("checkpoints/freevc-cvfr.pth", freevc_cvfr, None)
 
+print("Loading FreeVC-mls...")
+hps = utils.get_hparams_from_file("configs/freevc_nosr_mls.json")
+freevc_mls = SynthesizerTrn(
+    hps.data.filter_length // 2 + 1,
+    hps.train.segment_size // hps.data.hop_length,
+    **hps.model).to(device)
+_ = freevc_mls.eval()
+_ = utils.load_checkpoint("checkpoints/freevc-mls.pth", freevc_mls, None)
+
 print("Loading WavLM for content...")
 cmodel = WavLMModel.from_pretrained("microsoft/wavlm-large").to(device)
  
@@ -113,17 +122,19 @@ def convert(model, src_mic,src_file, tgt):
             audio = freevc_s.infer(c, mel=mel_tgt)
         elif model == "FreeVC CVFR":
             audio = freevc_cvfr.infer(c, g=g_tgt)
+        elif model == "FreeVC MLS":
+            audio = freevc_mls.infer(c, g=g_tgt)
         else:
             audio = freevc_24.infer(c, g=g_tgt)
         audio = audio[0][0].data.cpu().float().numpy()
-        if model == "FreeVC" or model == "FreeVC-s" or model == "FreeVC CVFR":
+        if model == "FreeVC" or model == "FreeVC-s" or model == "FreeVC CVFR" or model == "FreeVC MLS":
             write("out.wav", hps.data.sampling_rate, audio)
         else:
             write("out.wav", 24000, audio)
     out = "out.wav"
     return out
     
-model = gr.Dropdown(choices=["FreeVC", "FreeVC-s", "FreeVC (24kHz)","FreeVC CVFR"], value="FreeVC",type="value", label="Model") 
+model = gr.Dropdown(choices=["FreeVC", "FreeVC-s", "FreeVC (24kHz)","FreeVC CVFR", "FreeVC MLS"], value="FreeVC",type="value", label="Model") 
 audio1_mic=gr.Audio(source="microphone", type="filepath", label='record your voice', optional=True)
 audio1_file = gr.inputs.Audio(type='filepath', label='or upload an audio file', optional=True)
 audio2 = gr.inputs.Audio(label="Reference Audio", type='filepath')