Reproduce “Hierarchical Diffusion for Camera-controlled Autoregressive Video Generation” 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
Kaihua Chen, Yiming Gong, Deva Ramanan, Tarasha Khurana
The authors present WanAR + FixAR for generating long videos of static scenes from one image and a camera path. WanAR projects estimated scene structure into future views and inpaints missing regions; FixAR repairs sparse frames using longer-range context. On RealEstate10K, it reaches 18.31 mPSNR versus 16.89 for VACE, measuring co-visible-region alignment; depth errors and residual drift remain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.