Building on its recent progress in generative technology research, Meta has unveiled a new AI system that allows users to transform text prompts into short, high-quality video clips called Make-A-Video.
According to the tech giant’s blog post, the system “learns what the world looks like from paired text-image data and how the world moves from video footage with no associated text.” Meta noted that it plans to release a demo experience and will share details on the science in a cumulative research paper.
The system can essentially turn imagination into reality with one-of-a-kind videos full of colours, characters, landscapes and more. It can also craft videos from existing photos or take inspiration from old videos for fresh concepts. Notably, the feature utilises publicly available datasets.
Make-A-Video’s announcement comes shortly after the arrival of Meta’s Make-A-Scene system, which debuted earlier this year. As its name suggests, Make-A-Scene lets people create photorealistic illustrations and storybook-like art with text prompts, words and sketches.