Apr 23, 2026路also Affiliation:, D consistency. Vista, D-grounded priors for the video diffusion model. 3.2 Training with noisy multiview data So far, Eyeline Labs
Reshooting video from arbitrary viewpoints just got a whole lot better thanks to a 4D point cloud representation that maintains temporal consistency and precise camera control.
