The paper introduces Hyper-3DG, a method for generating 3D objects from textual descriptions. The method addresses the challenge of capturing high-order correlations between geometry and texture in 3D objects, which is often overlooked by existing methods. Hyper-3DG consists of two main components: the "Mainflow" and the "Geometry and Texture Hypergraph Refiner (HGRefiner)". The Mainflow uses a pre-trained 3D generator and a 2D diffusion model to initialize the 3D object. The HGRefiner refines the geometry and texture of the 3D object by patchifying it into smaller clusters and applying hypergraph learning to capture high-order correlations. The method improves the quality of 3D generation, particularly in terms of cross-view consistency, color and texture, and structural integrity, without increasing computational overhead. Experiments demonstrate the effectiveness of Hyper-3DG compared to state-of-the-art methods, and ablation studies highlight the importance of various components such as loss functions, hyperparameters, and pre-trained models. The method shows promise for applications in virtual reality, gaming, and architectural design.The paper introduces Hyper-3DG, a method for generating 3D objects from textual descriptions. The method addresses the challenge of capturing high-order correlations between geometry and texture in 3D objects, which is often overlooked by existing methods. Hyper-3DG consists of two main components: the "Mainflow" and the "Geometry and Texture Hypergraph Refiner (HGRefiner)". The Mainflow uses a pre-trained 3D generator and a 2D diffusion model to initialize the 3D object. The HGRefiner refines the geometry and texture of the 3D object by patchifying it into smaller clusters and applying hypergraph learning to capture high-order correlations. The method improves the quality of 3D generation, particularly in terms of cross-view consistency, color and texture, and structural integrity, without increasing computational overhead. Experiments demonstrate the effectiveness of Hyper-3DG compared to state-of-the-art methods, and ablation studies highlight the importance of various components such as loss functions, hyperparameters, and pre-trained models. The method shows promise for applications in virtual reality, gaming, and architectural design.