Stability AI has launched its latest AI model, Stable Virtual Camera, which transforms 2D images into "immersive" videos with realistic depth and perspective. Virtual cameras are commonly used in digital filmmaking and 3D animation, and with Stable Virtual Camera, Stability AI aims to integrate the power of generative AI, offering users greater control and customization.
Stable Virtual Camera can generate "novel views" of a scene from one or more images (up to 32), allowing users to specify camera angles. The model can generate videos with camera movement along dynamic or preset paths, including effects like spirals, zooms, pans, and dollys.
The current research preview version of Stable Virtual Camera generates videos in square (1:1), portrait (9:16), and landscape (16:9) aspect ratios, up to 1000 frames. However, Stability AI cautions that the model may produce lower-quality results in certain scenarios, particularly when processing images containing humans, animals, or "dynamic textures" like water.
Stability AI notes in its blog that "highly blurry scenes, complex camera paths intersecting objects or surfaces, and irregularly shaped objects may lead to flickering artifacts, especially when the target viewpoint differs significantly from the input images."
Currently, Stable Virtual Camera is available for research use under a non-commercial license on the Hugging Face platform. Stability AI is the company behind the popular image generation model Stable Diffusion. Despite past management issues and financial struggles, Stability AI secured new funding last year with investor support, aiming for a turnaround.
Recently, Stability AI appointed a new CEO and added film director James Cameron to its board. Furthermore, Stability AI has released several new image generation models and, in March of this year, partnered with chipmaker Arm to launch an AI model capable of generating audio and sound effects, aiming to bring these capabilities to mobile devices running Arm chips.
Key Highlights:
🌟 Stable Virtual Camera converts 2D images into immersive videos with various camera path options.
📉 The current model is a research preview and may produce lower quality results in some scenarios.
💼 After navigating management crises, Stability AI is actively restructuring and launching new products to improve its prospects.
