fastcore
cjm_transcription_plugin_system>=0.0.24
torch
numpy
soundfile
transformers
accelerate
librosa
mistral-common[audio]
uv
vllm
nest-asyncio
