How would you describe them?
[ECCV 2026] FixAnything: 3D-Consistent Rendering Refinement via Video Generative Priors
