Hello Creators,
Imagine transforming a single 2D image into an immersive 3D video, complete with realistic depth and perspective — all without the need for complex reconstruction or scene-specific optimization. Stability AI’s latest innovation, Stable Virtual Camera, currently in research preview, makes this possible.
What is Stable Virtual Camera?
Stable Virtual Camera is a multi-view diffusion model designed to generate 3D videos from one or more input images. It offers precise, intuitive control over 3D video outputs by allowing users to define camera trajectories and select from 14 dynamic camera paths, including 360°, Lemniscate (∞ shaped path), Spiral, Dolly Zoom, Move, Pan, and Roll.
Key Features
- Dynamic Camera Control: Users can specify custom camera movements or choose from predefined paths to navigate through the generated 3D scene.
- Flexible Inputs: The model can create 3D videos from a single image or up to 32 images, adapting to the available input to produce consistent and smooth outputs.
- Multiple Aspect Ratios: It supports various video formats, including square (1:1), portrait (9:16), landscape (16:9), and other custom aspect ratios without additional training.
- Long Video Generation: Stable Virtual Camera ensures 3D consistency in videos up to 1,000 frames, enabling seamless loops and smooth transitions, even when revisiting the same viewpoints.
Pushing the Limits of AI-Generated 3D Video
Stable Virtual Camera isn’t just another AI video tool — it’s setting new benchmarks in Novel View Synthesis (NVS). Compared to models like ViewCrafter and CAT3D, it delivers sharper, more realistic multi-view results. Whether handling large viewpoint shifts (big camera movements) or small viewpoint refinements (ensuring smooth transitions), it keeps everything looking cohesive and natural. This means no weird glitches, no warping — just seamless AI-generated 3D motion.
Getting Started
Stable Virtual Camera is available for research use under a Non-Commercial License. Researchers and developers can access the model’s weights on Hugging Face and the code on GitHub. This release invites the research community to explore its capabilities and contribute to its development.
Stable Virtual Camera is pushing the boundaries of AI-driven video creation, giving you full control over 3D perspectives with just a single image. Whether you’re a filmmaker, artist, or just an AI enthusiast, this tool unlocks a new way to play with depth, motion, and storytelling — without the hassle of complex 3D modeling.
Ready to experiment?
KINOMOTO.MAG 