Reproduce “VIDSTAMP: A Temporally-Aware Watermark for Ownership and Integrity in Video Diffusion Models” 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
Mohammadreza Teymoorianfard, Siddarth Sitaraman, Shiqing Ma, Amir Houmansadr
VidStamp embeds frame-specific watermarks during video generation by fine-tuning a temporal video decoder, avoiding a separate post-generation pass. Each frame carries its own message, supporting provenance and localization of swaps, insertions, and drops. In the SVD comparison, it carries 768 bits per 16-frame video and has a lower log P-value than VideoShield, with quality close to unwatermarked SVD; targeted removal remains untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.