Skip to main content
Parakeet MLX Hero Light

​
What is Parakeet MLX?

Parakeet MLX is a high-performance implementation of Nvidia’s Parakeet Automatic Speech Recognition (ASR) models, optimized specifically for Apple Silicon using the MLX framework. It brings state-of-the-art speech recognition capabilities to Mac users with native hardware acceleration.

Quick Start

Get started with a working transcription in under 2 minutes

Installation

Install Parakeet MLX with pip, uv, or as a CLI tool

Python API

Learn how to integrate Parakeet MLX into your Python applications

CLI Reference

Explore all command-line options and examples

​
Key Features

Parakeet MLX supports all major Parakeet model architectures:
  • TDT (Token-and-Duration Transducer): Best accuracy with beam search support
  • RNNT (RNN-Transducer): Efficient streaming transcription
  • CTC (Connectionist Temporal Classification): Fast processing for real-time applications
  • TDT-CTC Hybrid: Combines strengths of both approaches
All models are available through the mlx-community on Hugging Face.
Built-in support for streaming transcription allows you to process audio in real-time:
with model.transcribe_stream(context_size=(256, 256)) as transcriber:
    for chunk in audio_chunks:
        transcriber.add_audio(chunk)
        print(transcriber.result.text)
Perfect for live transcription, voice assistants, and real-time captioning.
Get precise timing information for every word:
result = model.transcribe("audio.wav")
for sentence in result.sentences:
    for token in sentence.tokens:
        print(f"{token.text} [{token.start:.2f}s - {token.end:.2f}s]")
Ideal for subtitle generation, video editing, and accessibility tools.
Export transcriptions in various formats:
  • TXT: Plain text transcription
  • SRT: SubRip subtitle format with timestamps
  • VTT: WebVTT format for web video players
  • JSON: Structured data with full timing and confidence scores
Enable word-level highlighting in subtitles with the --highlight-words flag.
Native MLX integration provides:
  • Hardware-accelerated inference on M-series chips
  • Efficient memory usage with bfloat16 precision
  • Local attention mechanisms for long-form audio
  • Chunking support for files of any length
All processing happens locally on your device—no cloud API required.
Use Parakeet MLX your way:
  • CLI: Simple command-line interface for batch processing
  • Python API: Full programmatic control for integration
  • Streaming API: Real-time transcription with context management
  • Low-Level API: Direct access to model internals for research

​
Why Parakeet MLX?

Local Processing

All transcription happens on your device. Your audio never leaves your machine.

No API Costs

Free to use with no per-minute charges or usage limits.

Open Source

Apache 2.0 licensed. Inspect, modify, and contribute to the code.

​
Supported Models

Parakeet MLX works with any compatible model from the mlx-community/parakeet collection:
ModelSizeTypeBest For
parakeet-tdt-0.6b-v3600MTDTGeneral use, best accuracy
parakeet-tdt-1.1b1.1BTDTHigh accuracy, larger context
parakeet-rnnt-0.6b600MRNNTStreaming applications
parakeet-ctc-0.6b600MCTCFast real-time processing
The default model is mlx-community/parakeet-tdt-0.6b-v3, which provides an excellent balance of speed and accuracy for most use cases.

​
Quick Example

Here’s how simple it is to transcribe audio with Parakeet MLX: 
parakeet-mlx audio.mp3

​
Use Cases

Content Creation

Generate accurate subtitles for videos, podcasts, and recorded lectures

Accessibility

Create real-time captions for live events and video conferencing

Research

Transcribe interviews, focus groups, and recorded observations

Documentation

Convert voice memos and meetings into searchable text

​
System Requirements

Parakeet MLX requires an Apple Silicon Mac (M1, M2, M3, or later) running macOS 13.0 or newer.
  • Processor: Apple Silicon (M1 or later)
  • RAM: 8GB minimum, 16GB recommended for larger models
  • Python: 3.10 or higher
  • Additional: ffmpeg for audio format conversion (CLI only)

​
Next Steps

1

Install Parakeet MLX

Follow the installation guide to set up Parakeet MLX on your system.
2

Try the Quick Start

Run your first transcription with the quick start guide.
3

Explore the API

Dive deeper into the Python API or CLI reference.

​
Community and Support

GitHub Repository

Star the repo, report issues, and contribute code

Hugging Face Models

Browse available Parakeet models optimized for MLX

Build docs developers (and LLMs) love

Get started for freeTalk to us