A new controllable video generation framework, UniCaMo, tackles one of the hardest problems in diffusion-based video: keeping object motion and camera motion coherent at the same time. It does this by shaping the input noise itself with 3D-grounded track warping and spherical noise sampling, instead of adding extra control branches. The result is state-of-the-art quality and motion control on standard benchmarks.
arXiv cs.GR
arXiv cs.GR
1% match
With your permission, GameDev.net uses analytics cookies to understand how people use the platform. You can accept analytics or continue with necessary cookies only. Learn more
