torch>=2.0.0
transformers>=4.45.0
accelerate
encodec
librosa
soundfile
gradio>=4.0
numpy
pandas
huggingface_hub