app.py

import os
import shutil
import zipfile
import gradio as gr


os.system('gdown https://drive.google.com/uc?id=1Flw6Z0K2QdRrTn5F-gVt6HdR9TRPiaKy')
with zipfile.ZipFile('VQMIVC-pretrained models.zip', 'r') as zip_ref:
    zip_ref.extractall('.')
    
shutil.move('VQMIVC-pretrained models/checkpoints/', '.')
shutil.move('VQMIVC-pretrained models/vocoder/', '.')



def inference(audio1, audio2):
  
  os.rename(audio1.name, '1.wav')
  os.rename(audio2.name, '2.wav')
  os.system('ls')
  os.system("python convert_example.py -s 1.wav -r 2.wav -c converted -m 'checkpoints/useCSMITrue_useCPMITrue_usePSMITrue_useAmpTrue/VQMIVC-model.ckpt-500.pt'")
  out = "converted/1_converted_gen.wav"
  return out

inputs = [gr.inputs.Audio(label="Source Audio", type='file'),gr.inputs.Audio(label="Reference Audio", type='file')]
outputs =  gr.outputs.Audio(label="Output Audio", type='file')


title = "VQMIVC"
description = "Gradio demo for VQMIVC: Vector Quantization and Mutual Information-Based Unsupervised Speech Representation Disentanglement for One-shot Voice Conversion. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below."
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2106.10132' target='_blank'>VQMIVC: Vector Quantization and Mutual Information-Based Unsupervised Speech Representation Disentanglement for One-shot Voice Conversion</a> | <a href='https://github.com/Wendison/VQMIVC' target='_blank'>Github Repo</a></p>"

examples=[['source.wav','ref.wav']]

gr.Interface(inference, inputs, outputs, title=title, description=description, article=article, examples=examples, enable_queue=True).launch()