GET3D is a generative model based on image learning that can directly generate 3D models with complex topology, rich geometric details, and high-fidelity textures. Trained on a 2D image collection by combining differentiable surface modeling, differentiable rendering, and 2D Generative Adversarial Networks, GET3D can generate high-quality 3D texture models covering various forms, including cars, chairs, animals, motorcycles, and characters, significantly improving upon previous methods.