transformers
gradio
sounddevice
numpy
pandas
speechbrain