Existing 3D generation frameworks have demonstrated notable success, their continued evolution is impeded by issues including view inconsistency and insufficient detail, which significantly hinders further methodological refinements. To address these challenges, we propose MagicGS++, an innovative framework for Image-To-3D generation that efficiently generates high-quality 3D content from a single-view image. MagicGS++ adopts a Coarse-To-Fine framework consisting of two main stages. In the frst stage, a fixed-orthogonal view diffusion model with a further Super-Resolution model are employed to generate orthogonal view images and perform initial optimization on the 3D Gaussian distribution, yielding a rough shape. In the second stage, a fixed view image of the 3D Gaussian model is rendered alongside two neighboring-view images most relevant to the fixed view. By jointly optimizing the multi-MSE loss between the neighboring view images and the orthogonal view image, and the SDS loss between the fixed-view image and the 3D model. Extensive experiments demonstrate that MagicGS++ outperforms existing methods in both geometric detail and texture quality. Moreover, by integrating with a conditional diffusion model, our method supports multi-modal 3D generation tasks, showcasing remarkable adaptability.

Recent advancements in 3D generation have garnered considerable interest due to their potential applications. Despite these advancements, the field faces persistent challenges in multi-conditional control, primarily due to the lack of paired datasets and the inherent complexity of 3D structures. To address these challenges, we introduce ImageBind3D, a novel framework for controllable 3D generation that integrates text, hand-drawn sketches, and depth maps to enhance user controllability. Our innovative contribution is adopting an inversion-align strategy, facilitating controllable 3D generation without requiring paired datasets. Firstly, utilizing GET3D as a baseline, our method innovates a 3D inversion technique that synchronizes 2D images with 3D shapes within the latent space of 3D GAN. Subsequently, we leverage images as intermediaries to facilitate pseudo-pairing between the shapes and various modalities. Moreover, our multi-modal diffusion model design strategically aligns external control signals with the generative model's latent knowledge, enabling precise and controllable 3D generation. Extensive experiments validate that ImageBind3D surpasses existing state-of-the-art methods in both fidelity and controllability. Additionally, our approach can offer composable guidance for any feed-forward 3D generative models, significantly enhancing their controllability.