1 paper
Dicong Qiu, Zhiyuan Xu, Yaosheng Liu +2
Modern video models can generate realistic videos from real appearance references and proxy renders that specify scene structure, viewpoint changes, and motion. Evaluating this ren…