RunwayML introduced a new text-to-video feature on its AI-backed web video editor. The new update can edit videos with written commands called “prompts.” Runway posted a video demonstration reel to show how a simple text input can alter a video. You write commands like “remove the object,” “make it more cinematic,” etc., in an input text box, and it gets done.
The promotional video also shows text-to-image generation similar to Stable Diffusion, text overlay, character masking, and much more.
Read More: AI & Us: a new documentary by Dataiku
Text-to-video technologies are yet in a primitive state because of high computational requirements and dataset scarcity for training with metadata. Runway’s text2video is not the first of its kind, there have been some other successful attempts like CogVideo. CogVideo can generate videos, however, with low resolution.
With every other attempt, it will be reasonable to expect enhanced quality in text2video technologies; hence, Runway’s teaser is a step forward in synthetic video generation.
Until now, Runway has been accessible as a web-based commercial video editing product. It is not entirely free and charges a monthly fee, including charges for cloud storage. Currently, the new text-to-video feature is in a controlled phase, with “Early Access” only available to a few people. You can sign up for the waitlist on Runway.