torch>=2.1
torchaudio>=2.1
numpy
safetensors
soundfile
huggingface_hub