Runway unveils Solaris, an AI model that generates interactive interfaces without code

Runway has unveiled Solaris, a new category of AI known as an ‘Interface World Model’ that generates interactive software interfaces in real time without the use of underlying code. Unlike traditional software that relies on pre-written logic and rendering engines, Solaris synthesizes visual frames one by one in response to user actions such as clicks, drags, or voice commands.

Technically, the system combines a language model to determine how an interface should evolve with a video generation model—built on Runway’s Gen-4 foundation—to render the visual output at 720p resolution. This approach treats the application as a continuous, responsive video stream rather than a fixed set of programmed elements.

Demonstrations of the technology include highly adaptive user experiences, such as a virtual clothing try-on where users drag garments onto a photo of themselves, and interactive simulations like a combustion engine or a customizable salad builder. While the technology currently faces challenges such as text garbling and session drift, Runway suggests it could eventually replace the concept of fixed ‘apps’ with dynamic, on-demand digital environments.