Reproduce “FlashRelight: Portrait Video Relighting with Dynamic Lighting” 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
Jiahui Sheng, Le Li, Limin Lin, Jiahao Li, Wansen Feng, Hong Gu, Shuhan Chen, Xiaorun Li and 1 more
FlashRelight relights portrait videos with time-varying illumination by first suppressing source lighting to create an approximate appearance guide, then generating the new lighting jointly. It supports environment-map sequences, text, and reference-video controls, though general text prompting is unreliable. On 10 held-out subjects and 16 unseen environment-map sequences, it reports lower lighting error and temporal LPIPS than DiffusionRenderer, only for this test.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.