Real-time face puppetry for the ComfyUI crowd

What it does AdvancedLivePortrait is a set of ComfyUI custom nodes that bolt facial expression editing onto the LivePortrait model. You can retarget a photo’s face, drive a video with new expressions, or extract and bank expression data from sample images. It also supports face tracking on a source video directly, so you can generate animation without a separate driving clip.

The interesting bit The author squeezes a real-time preview out of the pipeline and adds saveable expression data—essentially letting you build a library of reusable facial motions instead of one-off transfers. A motion_link system in the advanced workflow also lets you chain expressions with timed holds, turning the node graph into a crude but functional animation sequencer.

Key highlights

• Real-time preview, which the README claims is faster than the baseline LivePortrait experience
• Load Exp Data and Save Exp Data nodes that write expression banks to \ComfyUI\output\exp_data\
• A “video without a video” mode that tracks the source video’s own face rather than requiring an external driver
• Sequenced motion control via frame-length and wait-time indices in workflow2_advanced.json
• Registered with ComfyUI-Manager for plug-in installation

Caveats

• Documentation is thin: motion indices and node behavior are explained in a single screenshot and one equation, so plan on some guessing
• The project is largely glue and workflow logic around kijai’s model conversion, not a novel architecture

Verdict Grab it if you want node-level control over facial animation inside ComfyUI. If you’re looking for a standalone app or a new foundation model, keep walking.