Stability AI has unveiled the beta version of the Stable Video Diffusion video generator

Stability AI

Stability AI has announced Stable Video Diffusion, a video generator that allows you to create videos from a single image. The tool is available in test mode.

Here's What We Know

Stable Video Diffusion consists of two image-to-video models. They are capable of creating clips 14-25 frames long at 3-30 fps at a resolution of 576×1024 pixels.

According to the developers, Stable Video Diffusion is superior to counterparts from startups like Runway and Pika Labs.

The current version has a number of limitations: short duration, insufficient photorealism, lack of camera motion and text control. The system also doesn't handle people's faces well so far.

To train Stable Video Diffusion, open video data with millions of clips was used. This is an important point after the recent scandal around the illegal borrowing of content from Getty Images.

Potential users can apply for access to beta testing. The developers see prospects for applying the system in marketing, education and other areas.

Source: Stability AI