Favicon of MLX Audio

MLX Audio

A local AI audio library for Apple Silicon that generates and transcribes speech, processes audio, and provides an OpenAI-compatible API.

Screenshot of MLX Audio website

MLX Audio is a Python library for developers building speech and audio applications on Apple Silicon Macs. It runs models locally using Apple's MLX framework, with support for generated speech, transcription and audio processing.

  • Speech generation: Generate multilingual speech with Kokoro, Qwen3-TTS, Voxtral TTS and other models, with voice cloning, speaking speed control or voice design depending on the model.
  • Transcription and speaker labels: Transcribe recordings with Whisper, Parakeet or Qwen3-ASR, use Voxtral Realtime for streaming transcription, and get word-level timestamps or speaker labels with supported models.
  • Audio cleanup and separation: Extract sounds with text prompts through SAM-Audio, separate dialogue into speaker tracks with DialogueSidon, or reduce noise with MossFormer2 and DeepFilterNet.
  • Voice conversations: Build spoken interactions with Liquid2.5-Audio or use NemotronLabs VoiceChat for full-duplex conversations with streaming transcription and function calling.
  • Music generation: Create songs with multilingual lyrics using MiniMax Music 3.

The locally running REST server accepts OpenAI-compatible clients, and its web interface includes audio visualization. Quantized models can reduce model size and speed up inference on Apple Silicon. Language coverage, voice controls and streaming capabilities depend on the model you choose.

MLX Audio is open source under the MIT license. The companion mlx-audio-swift package provides on-device text-to-speech for macOS and iOS apps.

Similar to MLX Audio