omnivoice
torch
torchaudio
gradio
numpy
soundfile
spaces