Parakeet-MLX Audio Transcription

This skill provides instructions for using parakeet-mlx to convert audio files to text using NVIDIA's Parakeet automatic speech recognition model optimized for Apple's MLX framework.

Overview

Parakeet-MLX brings NVIDIA's Parakeet ASR model to Apple Silicon, enabling fast, high-quality on-device speech-to-text conversion similar to Whisper but optimized for Apple's MLX framework.

Key Features:

• High-quality transcription with timestamps
• Fast processing (e.g., 1+ hour audio in ~53 seconds)
• On-device processing (no cloud API required)
• Output in SRT subtitle format
• No installation required - run directly with uvx

Basic Usage

Transcribe an Audio File