Runway has demonstrated Solaris, a research system that generates an interactive user interface frame by frame instead of executing a conventional application. The 720p output responds to clicks, drags and voice commands, allowing the screen itself to change around a user's request.
The company calls Solaris an Interface World Model. A language model determines how the interface should respond, while technology derived from Runway's Gen-4.5 video model renders the result. Demonstrations include dragging clothing onto a photograph for a virtual fitting, rearranging furniture with spoken commands and showing visual steps in a tutorial rather than returning a text answer.
That flexibility also removes the stability people expect from software. Runway acknowledges that readable text is still unreliable, long sessions and consistent behavior remain open problems, and a plausible-looking wrong result can mislead users. Solaris does not yet support assistive technology such as screen readers, making accessibility another fundamental gap rather than a finishing detail. The company is seeking launch partners and offers an early-access form, but describes the current system as research rather than a ready replacement for apps.