faster-whisper @ https://github.com/SYSTRAN/faster-whisper/archive/8f01aee36b562e6be537e0341cdd40dc8bed33a7.tar.gz
websockets
onnxruntime==1.16.0
numba
kaldialign
soundfile
ffmpeg-python
scipy
jiwer
evaluate
numpy<2
tiktoken==0.8.0
openai-whisper==20240930