Stable Video 3D
Generate 3D video from a text or image prompt
Listed in categories:
Video3D ModelingArtificial Intelligence
Description
SV3D is a latent video diffusion model for high-resolution image-to-multiview generation of orbital videos around a 3D object. It adapts image-to-video diffusion model for novel multiview synthesis and 3D generation, leveraging generalization and multiview consistency of video models with explicit camera control for NVS. It also includes improved 3D optimization techniques for image-to-3D generation.
How to use Stable Video 3D?
To use SV3D, simply input a single image and let the model generate novel multiview images and 3D models. Use explicit camera control for customized view synthesis and leverage the improved 3D optimization techniques for accurate 3D generation.
Core features of Stable Video 3D:
1️⃣
Multiview image generation
2️⃣
3D model generation
3️⃣
Novel view synthesis
4️⃣
Explicit camera control
5️⃣
Improved 3D optimization
Why could be used Stable Video 3D?
| # | Use case | Status | |
|---|---|---|---|
| # 1 | Generating multiview images from a single image | ✅ | |
| # 2 | Creating 3D models from images | ✅ | |
| # 3 | Enhancing novel view synthesis | ✅ | |
Who developed Stable Video 3D?
The makers of SV3D are Vikram Voleti, ChunHan Yao, Mark Boss, Adam Letts, David Pankratz, Dmitrii Tochilkin, Christian Laforte, Robin Rombach, and Varun Jampani. They have developed SV3D for state-of-the-art performance in novel view synthesis and 3D generation.