transformers>=4.35
torch>=2.1,<2.4
soundfile
librosa
gradio>=4.0
datasets
optimum[openvino]