TL;DR: Text-to-3D generation of scenes decomposed automatically into objects, using only a pretrained diffusion model.

TL;DR: Text-to-3D generation of scenes that are decomposed automatically into objects.