Reproduce “WorldForge V2: Eliciting Camera-Trajectory Alignment 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

Chenxi Song, Yanming Yang, Tong Zhao, Ruibo Li, Joey Tianyi Zhou, Chi Zhang

WorldForge V2 extends camera-controlled video generation with depth-based warped views used as selective guidance during denoising. It aligns visible and generated regions’ denoising states and adapts blocks with LoRA. In the DL3DV comparison, H3 reaches target-view PSNR 20.21 versus 17.41 for TrajectoryCrafter; this result is not a camera-accuracy measure, and errors can persist with poor geometry or complex motion.

Summary by alphaXiv.

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