RODIN Diffusion is an AI system that automatically generates highly detailed 3D digital avatars. The generated avatars can be freely viewed in 360 degrees with unprecedented quality. This model significantly accelerates the traditionally complex 3D modeling process, opening up new opportunities for 3D artists. RODIN Model Architecture: This 3D digital avatar diffusion model is trained to generate 3D digital avatars represented as neural radiance fields. We leverage cutting-edge generative technology (diffusion models) for 3D modeling. We employ a triplane representation to decompose the neural radiance field of the avatar, allowing explicit modeling through diffusion models and rendering it as images via volume rendering. The proposed 3D-aware convolutions bring the required computational efficiency while maintaining the integrity of 3D diffusion modeling. The entire generation process is a hierarchical one, using cascaded diffusion models for multi-scale modeling. Once the generation model is trained, it can control the digital avatar generation based on latent codes derived from input images, text prompts, or random noise. RODIN Diffusion model visualization. For more details, please refer to the paper Rodin: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion.