fastcore
cjm_transcription_plugin_system>=0.0.16
torch
numpy
soundfile
cjm_ffmpeg_utils>=0.0.5
transformers<5
accelerate
librosa
mistral-common[audio]
