builder/fetch_models.py

import se_extractor as se

_ = se.generate_voice_segments('openai_source_output.mp3',vad=False)
_ = se.generate_voice_segments('openai_source_output.mp3',vad=True)

from resemble_enhance.enhancer.inference import denoise, enhance
import torchaudio


dwav, sr = torchaudio.load('openai_source_output.mp3')
dwav = dwav.mean(dim=0)

wav1, new_sr = enhance(dwav, sr, 'cuda:0', nfe=32, solver='midpoint', lambd=0.9, tau=0.5)