README.md

---
title: OmniVoice MLX
emoji: 🌍
colorFrom: green
colorTo: indigo
sdk: gradio
sdk_version: 6.14.0
app_file: app.py
pinned: false
license: apache-2.0
short_description: OmniVoice text-to-speech demo using the MLX backend
models:
- mlx-community/OmniVoice
- mlx-community/OmniVoice-4bit
- mlx-community/OmniVoice-8bit
- mlx-community/OmniVoice-bfloat16
- mlx-community/OmniVoice-fp32
---

# OmniVoice MLX

This Space adapts the official [k2-fsa/OmniVoice Space](https://huggingface.co/spaces/k2-fsa/OmniVoice) to load the community MLX backend from [ailuntx/OmniVoice-MLX](https://github.com/ailuntx/OmniVoice-MLX).

The default model is `mlx-community/OmniVoice-4bit` to reduce memory pressure. Set `OMNIVOICE_MODEL` to another MLX repo if needed.

MLX performs best on Apple Silicon. Hugging Face Spaces run on Linux, so this demo uses the MLX Linux CPU backend and can be slow. For normal interactive use, install and run the model locally on an Apple Silicon Mac.