Researchers have developed a new watermarking technique called LoT-Pass, designed to trace the origin of images used in image-to-video generation. This method addresses the limitations of existing watermarking techniques that fail to maintain robustness across different modalities. LoT-Pass introduces the concept of Robust Diffusion Distance to measure temporal persistence and uses a cross-modal watermarking framework with a video-simulation noise layer and an optical-flow-based alignment module. Experiments show that LoT-Pass significantly enhances watermark robustness and imperceptibility in generated videos. AI
IMPACT Enhances the traceability of AI-generated video content, potentially mitigating misuse in misinformation.
RANK_REASON The cluster contains an academic paper detailing a new technical method for image watermarking. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- DagsHub
- Guanjie Wang
- Hugging Face
- I2VWM
- LoT-Pass
- Robust Diffusion Distance
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
