SV3D
Stability AI's image-to-3D model — 21-frame orbital videos from one image, research license.
About
Stable Video 3D (SV3D) is the novel multi-view synthesis model Stability AI released on March 18, 2024, built on the Stable Video Diffusion backbone. Given a single image of an object (ideally on a white background), it generates a 21-frame orbital video at 576×576 — effectively turntable views that can be lifted into 3D meshes via downstream reconstruction. Two variants shipped: SV3D_u for unconditioned orbital videos and SV3D_p, which accepts camera-path conditioning for 3D video along specified elevation/azimuth trajectories.
The honest picture: SV3D was released strictly for research purposes under Stability AI's non-commercial research community license — you cannot use it in commercial products. Single-image input quality dominates output quality (cluttered backgrounds degrade views fast), and the orbital videos still need a separate reconstruction step (NeRF, Gaussian splatting, or mesh extraction) before they become usable 3D assets. Within those bounds, it remains one of the cleanest open multi-view diffusion checkpoints available.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 113; Trending score 62/100 from researcher assessment.