Reproduce “PixReenact: Pixel-Conditioned Causal Video Diffusion for Streaming Head-Avatar Reenactment” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Gavriel Habib, Dvir Samuel, Or Shimshi, Rami Ben-Ari
PixReenact streams head-avatar animation by conditioning a causal video generator on VAE-encoded reference and driver frames, training it to preserve reference identity while following driver motion. Cross-identity supervision, corrective objectives, and local driver context support this separation. On 68 extreme-condition pairs, its reference-identity score is 0.727 versus PersonaLive’s 0.472, though PersonaLive scores better on pose distance and FVD.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.