Avatarify
Open-source real-time face animation that drives a still portrait from your webcam.
About
Avatarify is an open-source project for real-time face reenactment: point your webcam at it and your head movements and expressions are transferred live onto a still portrait, powered by the First Order Motion Model. The result pipes out through a virtual camera into Zoom, Teams, OBS or any video app, so you can appear as a character, celebrity lookalike or synthetic face in meetings and streams.
It runs on PyTorch with optional ONNX, OpenVINO and TensorRT backends, plus a remote-GPU mode for machines without a capable card, and an optional Wav2Lip integration adds lip-sync. The underlying models are research/non-commercial licensed, and real-time performance wants a decent GPU. Community forks keep the project alive for Windows, macOS and Linux.
What it can do
- Real-time face reenactment from webcam
- Virtual camera output to meeting apps
- Multiple inference backends
- Remote GPU mode via Colab/Kaggle
- Optional Wav2Lip lip-sync
Setup
Clone the repo and run the setup script, or use a community portable build.
Prerequisites
Python 3.10+, ideally an NVIDIA GPU; OBS Studio for virtual camera output.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 93; Trending score 68/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.