Images created using artificial intelligence Stable Diffusion can now be animated, the startup Stability AI has announced. The company unveiled a new product called Stable Video Diffusion, which allows users to create a video from a still image. The new feature has been made available in the form of two image-to-video converters, each capable of generating 14 to 25 frames at between 3 and 30 frames per second with a resolution of 576 x 1024 pixels. The tool is capable of synthesizing multiple views from one image.
Stable Video Diffusion is in this momentarily available for research purposes only, not for individuals or commercial entities. Those interested can register and join the waiting list. The tool will have potential uses in industries such as advertising, education or entertainment.
However, Stable Video Diffusion has some limitations. It generates relatively short videos (less than 4 seconds), lacks photorealism, cannot properly imitate camera movement, cannot generate readable text, and does not always generate persons and faces correctly. The tool was trained on a dataset containing millions of videos, with Stability AI saying it used videos that were publicly available for research purposes. The origin of the dataset is important given that Stability AI was recently sued by Getty Images for abusing its image archives in an artificial intelligence exercise.