Reproduce “SIGMark: Scalable In-Generation Watermark with Blind Extraction for Video Diffusion” 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
Xinjie Zhu, Zijing Zhao, Hui Jin, Qingxiao Guo, Yilong Ma, Yunhao Wang, Xiaobing Guo, Weifeng Zhang
SIGMark embeds blind watermarks in diffusion-generated video by encoding messages into generator noise with shared frame-wise pseudorandom error-correcting keys, then restoring temporal frame groups before extraction. On HunyuanVideo image-to-video, it reaches 0.905 bit accuracy for 8,192 bits, versus 0.999 for non-blind VideoShield. At 3,200 videos, in the reported experiment, extraction takes 14.2 seconds versus 107.7.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.