Instructions to use Avdpro/MLX-LivePortrait with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use Avdpro/MLX-LivePortrait with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir MLX-LivePortrait Avdpro/MLX-LivePortrait
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
MLX LivePortrait
Native MLX conversion of the official LivePortrait human portrait-animation checkpoint for AI2Apps on Apple Silicon.
The repository contains seven deterministic NPZ weight files and an oMLX YuNet
face-detector bundle. It excludes InsightFace weights. Use the signed
ai2apps/model-liveportrait-mlx Package for inference, provenance checks,
multipart input validation, progress, cancellation, and audio preservation.
Default precision is FP32; BF16 is an explicit faster profile. FP16 is not supported.
Hardware compatibility
Log In to add your hardware
Quantized