torch>=2.0
numpy
librosa
soundfile

[modelscope]
modelscope

[server]
fastapi
uvicorn
python-multipart
python-dotenv
