¹Google Research      ²VGG, University of Oxford ³Google DeepMind TL;DR: Feed-forward 3D scene generation in 6.25s on a single GPU. How it works Given one or more input images, we generate multi-view Splatter Images. To do so, we first generate the scene appearance and geometry using a multi-view diffusion model. Then, Splatter Images are regressed using a Gaussian Head. 3D Gaussians from multiple Splatter Images are combined to form the 3D scene. An animated diagram briefly describing the metho...