Stability AI Unveils Stable Virtual Camera for 2D-to-3D Video Generation

Stability AI has introduced Stable Virtual Camera, a new AI model designed to transform 2D images into immersive videos with realistic depth and perspective. The model generates “novel views” from one or more images (up to 32) and allows users to create dynamic camera movements like Spiral, Dolly Zoom, Move, and Pan.

Currently available as a research preview, Stable Virtual Camera can produce videos up to 1,000 frames long in various aspect ratios. However, Stability AI warns of potential quality issues with human and animal images or highly dynamic textures like water. The model is available for noncommercial research use on Hugging Face.

This release comes as Stability AI undergoes major restructuring following financial struggles. The company has recently appointed a new CEO, added filmmaker James Cameron to its board, and expanded its AI offerings, including a collaboration with Arm to bring AI-generated audio to mobile devices.