Reproduce “Coda-DaVinci: Reimagining Video Generation with Code2Video” 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
Yiyan Ji, Qixun Wang, Yang Shi, Zhuoran Zhang, Yifan Xu, Tengfei Liu, Yuqi Tang, Xiaohan Zhang and 15 more
Coda-DaVinci uses selective executable 3D scene proxies to represent layout, interactions, and persistent state, while leaving appearance and fine motion to the video model. It uses coarse human proxies or direct generation when detailed control could constrain natural movement. On VTR-Bench, it scores 91.42 in checklist satisfaction and 11.08% WER, but does not lead every metric.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.