Automatic 3D generation has recently attracted widespread attention. Recent methods have greatly accelerated the generation speed, but usually produce less-detailed objects due to limited model capacity or 3D data. Motivated by recent advancements in video diffusion models, we introduce \approach, which leverages the world simulation capacity of pre-trained video diffusion models to facilitate 3D generation. To fully unleash the potential of video diffusion to perceive the 3D world, we further introduce geometrical consistency prior and extend the video diffusion model to a multi-view consistent 3D generator. Benefiting from this, the state-of-the-art video diffusion model could be fine-tuned to generate 360-degree orbit frames surrounding an object given a single image. With our tailored reconstruction pipelines, we can generate high-quality meshes or 3D Gaussians within 3 minutes. Furthermore, our method can be extended to scene-level novel view synthesis, achieving precise control over the camera path with sparse input views. Extensive experiments demonstrate the superior performance of the proposed approach, especially in terms of generation quality and multi-view consistency. Our code is available at https://github.com/heheyas/V3D
Our approach is based on the following research or resources.
Stable Video Diffusion introduces an idea similar to our windowed position encoding for coarse-to-fine optimization.
threestudio provides a great baseline for text-to-3D generation.
3D Gaussian Splatting the pioneering work achieves superior performance and enables real-time rendering.
splat a WebGL based gaussian renderer.
@article{chen2024v3d,
author={Zilong Chen and Yikai Wang and Feng Wang and Zhengyi Wang and Huaping Liu},
title={V3D: Video Diffusion Models are Effective 3D Generators},
journal={arXiv preprint arXiv:2403.06738},
year={2024}
}