MLX STT

Speech-To-Text/ASR/Transcribe with MLX (Apple Silicon) and GLM-ASR-Nano-2512 locally.

Free and Accurate. No api key required. No server required.

Requirements

• mlx: macOS with Apple Silicon
• brew: used to install deps if not available

Installation

bash ${baseDir}/install.sh

This script will use brew to install these cli tools if not available:

• ffmpeg: convert audio format when needed
• uv: install python package and run python script
• mlx_audio: do the real job

Usage

To transcribe an audio file, run the mlx-stt.py script:

uv run  ${baseDir}/mlx-stt.py <audio_file_path>

• When first run, it will download model from Hugging Face, default: mlx-community/GLM-ASR-Nano-2512-8bit, 2.5GB ish.
• The transcript result will be printed to stdout.