VGGT-Edit: Peking University, CUHK & Shanghai AI Lab Jointly Debut 3D Scene Editing Framework at 120x Speed

A joint team from Peking University, CUHK, Shanghai AI Lab, and NTU introduced VGGT-Edit, a native 3D scene editing framework.

Unlike 2D view-propagation methods, it edits reconstructed scenes directly in 3D for better semantic consistency and multi-view stability.

On the DeltaScene benchmark, VGGT-Edit reportedly delivers strong results while cutting edit time to about 5 seconds.

The system could enable faster, more interactive 3D editing workflows for games, film, VR, and other content production tasks.