Stability AI Unveils Stable Virtual Camera for Easy 3D Video Generation
Stability AI releases Stable Virtual Camera, a new multi-view diffusion model that turns single 2D images into immersive 3D videos. The tool offers precise 3D camera controls and is currently available for non-commercial research use.
Stability AI introduces Stable Virtual Camera, a multi-view diffusion model currently in research preview that transforms 2D images into immersive 3D videos with realistic depth and perspective. Unlike traditional 3D video models that require large sets of input images or complex preprocessing, this new tool generates novel views of a scene without complex reconstruction or scene-specific optimization. The model builds upon the concept of a traditional virtual camera used in filmmaking by combining that familiar control with the power of generative AI.
The tool provides precise, intuitive control over 3D video outputs by supporting user-defined camera trajectories alongside 14 dynamic camera paths. These preset paths include popular cinematic movements such as 360°, Lemniscate, Spiral, Dolly Zoom, Move, Pan, and Roll. Additionally, the system features flexible input options that allow users to generate 3D videos from just a single image or up to 32 images, producing consistent and smooth trajectory videos across multiple aspect ratios.
Stability AI makes Stable Virtual Camera available to the research community under a Non-Commercial License to encourage exploration and contribution to its development. Researchers can access the technical details by reading the official paper, download the model weights on Hugging Face, and review or utilize the source code directly on GitHub. This release represents a significant step forward in making high-quality 3D video generation more accessible and controllable for AI researchers.