Runway's Solaris generates an app interface frame by frame, with no code

Runway’s newest model does not generate a video of an app. It generates the app.

What Solaris is

Introduced on August 31, 2026, Solaris is the first of what Runway calls Interface World Models. Instead of code that describes buttons and screens, a single world model generates every frame of the interface and every response to user input, live, at 720p and interactive speeds. A language model sits alongside it and decides how the interface should evolve when the user clicks, types or asks for something; the world model renders the result. The system is built on Runway’s Gen-4.5 video generation model.

The study

Runway ran a 250-person study across 30 interaction examples, collecting nearly 7,500 pairwise judgments. Participants preferred Solaris over conventionally coded interfaces 61% to 24% on following instructions, and 71% to 21% on behaving naturally. Solaris is not public: Runway says it is working with partners on a launch and takes early-access requests through a form.

What to watch

A painted interface has obvious costs, from latency to the impossibility of guaranteeing that a button does the same thing twice. It also has an obvious appeal: an interface that reshapes itself around what the user is trying to do, without a developer writing the screen first. The preference numbers come from Runway’s own study, so the real test is what partners ship with it and how the frame-by-frame approach holds up under ordinary, repetitive software work.