Runway Gen-2 reveals limitations of current text-to-video technology

Artificial Intelligence (AI) start-up Runway has released its Gen-2 model, a commercially available application that generates videos from text prompts or existing images. Using diffusion technology, it learns how to gradually reduce the noise within an image to a point closer to a prompt chosen by the user. The new version provides four-second-long videos, which some reviewers have criticised for not being of high enough quality. However, Runway CEO Cristóbal Valenzuela sees the technology as a useful tool for artists and designers rather than a fully operational video editing system. Despite Gen-2’s limitations, it represents a step forward towards the creation of AI that can produce fully-fledged movies.