Reproduce “One Image Is Enough: Self-Generated Supervision for Personalizing Human Video Diffusion at Test Time” 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
Sadman Rakib Pinon, Adamo Orsini, Yanis Falaki, Linlian Jiang, Jacob Kingi, Yang Wang, Xinxin Zuo
The method personalizes pose-driven human-video generation from a reference portrait without footage of the test subject. It generates pose-specific images, selects them for resemblance to the portrait, and adapts the animator using repeated still clips. On TikTok, it scores 133.5 FVD versus 157.3 for Real-5, with FaceID 0.617 versus 0.615; BodySIM is slightly lower, 0.887 versus 0.888.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.