Vista4D: Video Reshooting with 4D Point Clouds, CVPR 2026 Highlight

rsasaki0109 · x · 2026-08-14

Vista4D is a video reshooting framework that synthesizes dynamic scenes from novel camera trajectories and viewpoints. It bridges the distribution shift between training and inference for point-cloud-grounded video reshooting by training on noisy, reconstructed multiview videos, making it robust to point cloud artifacts. Its 4D point cloud with temporally-persistent static points preserves scene content and improves camera control. It generalizes to dynamic scene expansion, 4D scene recomposition, and long video inference.

Original post →

More from Research

Research channel →