How AI describes CLAY

CLAY is a large-scale generative model that creates high-quality 3D assets by turning text, images, or 3D-aware inputs into detailed geometry and physically-based textures. It combines a multi-resolution VAE with a latent Diffusion Transformer to learn rich 3D priors from diverse geometries and represents surfaces with neural fields, enabling a transformer-based latent-space geometry generator. It supports multiple conditioning modalities (multi-view images, voxels, bounding boxes, point clouds, implicit representations) and outputs production-ready assets with 2K PBR textures for realistic rendering.