# Blaizzy/mlx-audio

A text-to-speech (TTS), speech-to-text (STT) and speech-to-speech (STS) library built on Apple's MLX framework, providing efficient speech analysis on Apple Silicon.

Repository: https://github.com/Blaizzy/mlx-audio
Canonical: https://ross.abutalabs.com/products/mlx-audio
Homepage: https://blaizzy.github.io/mlx-audio/
Language: Python
License: MIT
License Family: permissive
Topics: apple-silicon, audio-processing, mlx, multimodal, speech-recognition, speech-synthesis, speech-to-text, text-to-speech, transformers
Last push: 2026-08-26T18:40:08+00:00

## Health v2 (maintenance only)
Score: 88/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 98, longevity 46
- inputs: {"age_days": 644, "days_push": 7, "days_rel": 16, "gap_med": 14, "n_releases_24m": 28}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7793, forks 698 (observed 2026-08-28T04:10:05.046706+00:00)

## What it is
MLX-Audio is a Python library built on Apple's MLX framework for fast text-to-speech (TTS), speech-to-text (STT), and speech-to-speech (STS) inference on Apple Silicon Macs. It supports multiple model architectures, voice cloning, quantization, a CLI, an OpenAI-compatible REST API server with a web UI, and a Swift package for iOS/macOS.

## Use cases
- generate speech from text on my mac
- transcribe audio files locally with whisper
- clone a voice for text-to-speech
- run a local openai-compatible tts api server
- convert text to speech on apple silicon
- stream speech-to-text with word-level timestamps
- on-device tts for ios app
- separate vocals from background audio

## When to choose
- you want fast local speech inference on Apple Silicon M-series chips
- you need TTS, STT, or STS with voice cloning and multilingual support
- you want an OpenAI-compatible audio API server running locally
- you need quantized models for memory-efficient inference on a Mac

## When to avoid
- you need GPU inference on NVIDIA/AMD hardware or Linux servers
- you need production-grade enterprise speech services with SLAs
- you work primarily on non-macOS platforms
- you need training or fine-tuning of speech models rather than inference

## Facets
- artifact type: library
- maturity: active
- function: tts, speech-recognition, audio-processing, llm-inference, cli, http-server
- domain: speech-processing, machine-learning, artificial-intelligence
- platform: python, cli
- tags: apple-silicon, mlx, voice-cloning, speech-to-speech, quantization, openai-compatible-api, music-generation, audio, macos, swift, web-server

## Member repositories
- Blaizzy/mlx-audio (main) score 88

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:05.046706+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:34:38.570073+00:00, confidence not recorded.
  - readme: https://github.com/Blaizzy/mlx-audio (fetched 2026-08-28T04:10:05.046706+00:00, sha a07651feffae)
  - homepage: https://blaizzy.github.io/mlx-audio/ (fetched 2026-08-29T08:30:52.367685+00:00, sha 0b1315aa9c32)
  - registry_pypi: https://pypi.org/pypi/mlx-audio/json (fetched 2026-08-29T08:30:52.376793+00:00, sha a42ef29ca056)
- Data as of 2026-08-30T08:39:29.467469+00:00.