Alibaba presents the RichDreamer model, a Normal-Depth diffusion model for detail richness in text-to-3D. The model addresses the key challenge of lifting 2D diffusion to 3D generation by learning a joint distribution of normal and depth information. It shows superior performance in geometry and appearance modeling compared to existing approaches.
