Recent compelling successes in 2D diffusion models and large vision language models provide new possibilities for generating 3d assets.

In this work, we introduce Wonder3D, a novel method for efficiently generating high-fidelity textured meshes from single-view images.