We introduce a native 3D generative model scalable to in-the-wild input images, including D3D-VAE, a novel 3D variational auto-encoder effectively encoding a 3D point cloud into a triplane latent, and D3D-DiT, a scalable image-conditioned 3D diffusion transformer capable of generating 3D asserts consistent with input images.

Generating high-resolution 3D shapes using volumetric representations such as Signed Distance Functions (SDFs) presents substantial computational and memory challenges. We introduce Direct3D-S2, a scalable 3D generation framework based on sparse volumes that achieves superior output quality with dramatically reduced training costs......