whispermlx is a command-line tool for transcribing audio files on Apple Silicon. It runs the full three-stage pipeline — VAD, Whisper transcription, optional alignment, and optional diarization — from a single command.
Basic syntax
Quick examples
Check version
Audio files
The first positional argument (or multiple) is the audio file path(s) to transcribe. You can pass any number of files and whispermlx will process them in order.ffmpeg can decode.
MLX Whisper inference always runs on the Apple Silicon GPU automatically. The
--device option only controls the device used for VAD, alignment, and diarization models.