Stable Video 3D
Reconstructs multi-view and 3D shape of an object from a single image
WHAT IT IS
Stable Video 3D (SV3D) is a model Stability AI released in March 2024 with open weights. Given a single image, it generates a sequence of views orbiting the object and can export a textured 3D mesh directly. It builds on Stable Video Diffusion’s image-to-video base, turning video generation toward novel-view synthesis. It shipped in two versions, SV3D_u without camera conditioning and SV3D_p with orbital camera conditioning.
Why it matters
It connects image-to-video capability to 3D reconstruction, reusing single-image video generation as orbital view synthesis — a simple route within open-source 3D generation.
Key specs
- Input
- Single image
- Output
- Multi-view images and 3D mesh
- Versions
- SV3D_u and SV3D_p
- Open weights
- Yes
- Released
- 2024-03
Capabilities
Related concepts
Diffusion Models
Learn a thousand tiny denoising steps, and you can build an image from pure noise
Generative Models: An Overview
Discriminative models answer "what is this"; generative models answer "what should this look like"
Image Representation
To a machine, a photo is nothing but stacked grids of numbers