# Optional: keep the background (music, room, effects) under a translated dub. Demucs separates the voice from the rest
# (separate.py runs it in a child process). Pure additions to the torch that requirements-torch.txt installed; the htdemucs
# weights (about 80 MB) come from the Hugging Face hub on first use. Without it a dub simply drops the whole sound while a line is spoken.
demucs==4.1.0
