Reproduce “From Short Clips to Real-Time Streams: Identity-Aware Supervision and Context-Efficient Generation for Video Face Swapping” 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

Zekai Luo, Yuwei Liang, Ruiqi Ma, Haoyu Huang, Xuehao Wang, Huanyi Zheng, Mingyu Liu, Linhao Zhong and 3 more

StreamingSwap enables face swapping by supervising source identity, generating within a face region, and continuing in short windows with recent outputs as history. On 200 long videos, it reports 21.174 newly generated frames per second on one RTX 4090. Against DreamID-V, it has lower ID similarity but better lighting error and FVD; this is a trade-off, not leadership.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.