transcribe

Transcribe a video or audio file to text, locally, using Whisper on Apple Silicon (mlx-whisper).

Getting Started

Prerequisites

• macOS on Apple Silicon
• ffmpeg — brew install ffmpeg
• uv — brew install uv (manages Python and dependencies automatically)

Usage

Drop video/audio files into the temp/ folder, then run:

uv run transcribe.py

Every media file in the folder is transcribed; files that already have a transcript are skipped. You can also pass a specific file or a different folder:

uv run transcribe.py path/to/video.mp4

The transcript is written next to the input file (path/to/video.txt).

The first run downloads the Whisper model (~1.6 GB) from Hugging Face; later runs use the cached copy. To trade accuracy for speed, pass a smaller model:

uv run transcribe.py path/to/video.mp4 --model mlx-community/whisper-small

License

See LICENSE.