torch
torchvision
torchaudio
numpy<2.0
opencv-python
librosa==0.9.2
scipy
ffmpeg-python
gradio
tqdm
soundfile
huggingface_hub