Stable Diffusion
Open models in every modality, for everyone, everywhere.
4.8•27 reviews•278 followers
Open models in every modality, for everyone, everywhere.
4.8•27 reviews•278 followers
This is the 3rd launch from Stable Diffusion. View more
Stable Virtual Camera
Introducing Stable Virtual Camera, currently in research preview. This
multi-view diffusion model transforms 2D images into immersive 3D videos with realistic depth and perspective—without complex reconstruction or scene-specific optimization.
Free
Launch Team
Hi everyone!
Stability AI has a research preview out – Stable Virtual Camera – that turns a single photo (or a few) into a 3D video with impressive camera control!
You provide one or more images of a scene, and then specify how you want the camera to move (using presets like "zoom" or "orbit," or even defining a custom path). The model then generates a 3D-consistent video following that trajectory.
It's built on diffusion models, but it's not just generating pixels. It's creating a coherent 3D representation, which allows for smooth, long videos (up to 1000 frames!). They've achieved state-of-the-art results on novel view synthesis benchmarks.
The code is open-source, though the model itself is under a non-commercial license for now. While it works well in many cases, they do note that it can struggle with humans, animals, and dynamic textures like water in its current version. Very honest of them.
Give it a try – upload an image here and see what it can do.