Parakeet MLX for Openclaw

A high-performance command-line tool for private, local speech-to-text transcription optimized for Apple Silicon.

kylehowells
v1.0.0
Jan 13, 2026
2
2.7k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install parakeet-mlx

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install parakeet-mlx using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Parakeet MLX?

Parakeet MLX is a powerful Automatic Speech Recognition (ASR) tool specifically designed to leverage the power of Apple Silicon. By utilizing the MLX framework, it provides fast and accurate transcription services entirely on-device, removing the need for cloud-based APIs or expensive subscriptions. It is a vital component for developers building local-first workflows within the Openclaw Skills framework.

The tool is built to handle various audio formats and can output transcriptions in multiple industry-standard formats. Because it runs locally, users maintain full control over their data privacy while benefiting from the hardware acceleration provided by the M-series chips.

Parakeet MLX Use Cases

  • Transcribing private meetings or interviews without sending data to third-party servers.
  • Generating SRT or VTT subtitles for video content directly on a Mac.
  • Batch processing large libraries of audio files using shell wildcards for automation.
  • Integrating offline speech-to-text capabilities into local AI agent workflows via Openclaw Skills.

How Parakeet MLX Works

  1. The user executes the command pointing to one or more local audio files.
  2. On the first run, the tool automatically downloads the optimized Parakeet TDT model from Hugging Face to a local cache.
  3. The system utilizes ffmpeg to decode the audio stream and prepare it for the model.
  4. Transcription is performed locally using the MLX framework, which is specifically tuned for Apple Silicon GPUs and Neural Engines.
  5. The tool generates output files in the requested formats (txt, json, srt, etc.) in the same directory or a specified path.

Parakeet MLX Setup

To get started with this tool as part of your Openclaw Skills setup, ensure you have ffmpeg installed on your system. Then, install the CLI using the uv package manager:

uv tool install parakeet-mlx -U

You can verify the installation and see all available configuration options by running:

parakeet-mlx --help

Parakeet MLX Data Schema & Taxonomy

The tool manages data through local file outputs and a specific cache directory for models. Below is the organizational structure:

Data Type Location / Format Details
Models ~/.cache/huggingface Stores the downloaded MLX-optimized weights.
Transcription .txt, .srt, .vtt Standard text and subtitle output files.
Metadata .json Includes timestamps, segments, and confidence scores.
Audio Input .mp3, .m4a, .wav, etc. Supports any format compatible with ffmpeg.

Parakeet MLX Advanced Features

  • Word-level highlighting for precise subtitle creation using the --highlight-words flag.
  • Simultaneous generation of all supported output formats with the --output-format all command.
  • Verbose mode for real-time progress monitoring and detailed confidence scoring per segment.
  • Native support for shell wildcards, allowing for the transcription of entire directories in a single command.
  • Optimized for the mlx-community/parakeet-tdt-0.6b-v3 model for the best balance of speed and accuracy on Mac hardware.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*