Stable Video 3D
Stable Video 3D adapts Stable Video Diffusion to generate orbital views of an object from one image. The release contains an unconditioned-orbit variant and a camera-conditioned variant.
Model design and access
The model was fine-tuned from SVD using rendered Objaverse objects. Its immediate output is a sequence of object views. A view sequence and a finished mesh are different artifacts; reconstruction requires a further process.Stable Video 3D official model card
Practical use
Practical guidance. SV3D can provide view proposals for asset reconstruction or turntable previews. Use a clean object image and inspect the hidden side that the model invents. Camera conditioning controls the requested views, but it cannot recover true geometry that is absent from the source image.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | SV3DSV3D_uSV3D_p |
| Output | 21 frames at 576 × 576.Stable Video 3D official model card |
| Variants | SV3D_u uses an image; SV3D_p also accepts camera-path conditioning.Stable Video 3D official model card |
| License | Stability AI Community License.Stable Video 3D official model card |
Related models
- Stable Fast 3D
Stability AI · Stable 3D
Stable Fast 3D reconstructs a textured mesh from a single object image.
Launch date not confirmedAPI · Downloadable - Stable Video 4D
Stability AI · Stable Video
Stable Video 4D generates novel views of a moving object from a single-view video.
Launch date not confirmedDownloadable