Runway Solaris: A real-time interface – and why it’s not a replacement for web development

Runway introduced Solaris, an early research system from the Interface World Models family. It does not assemble an application from components and code: the model generates a picture of the interface and its response to user action in real time. This is an interesting shift for visual prototypes and immersive brand scripts, but not a ready-made replacement for a conventional app.

>

What did Runway say? h2 >>

An ordinary digital product passes the chain “makeup → code → interface”. In Solaris, the visual scene itself becomes an interface: the user clicks, drags an object or writes a request, and the model synthesizes the next frame and reaction of the scene. Runway describes this as a system where the language model interprets intent and sets behavior, and the visual model continuously renders the result. </ p. >>

In the demo, this looks convincing for tasks where space, material, movement and direct interaction with the object are important: a virtual showcase, fitting the subject, learning with an interactive scene. For the designer, what is valuable here is not the “magical generation of the site”, but the ability to quickly check how the user perceives and changes the visual idea before expensive development.

>

Why this is not a reason to throw out Figma and frontend h2 >>

The most important limitation Runway calls itself. Stable readable text remains a challenge for video generation. The model also does not guarantee the reliability of commercial or educational information, the stability of a long session, work with screen readers and integration with the rest of the software stack. To generate each frame is more expensive than once to give the collected page. >

Therefore, Solaris cannot be considered a technology for payment, personal account, form with legally relevant data or interface, which must be accessible and reproducible. A beautiful response model does not equal the correct state of the product. In such scenarios, deterministic components, data, and availability remain mandatory.

>

Where to experiment now

>

>
• Concept and pitch: to turn a static art direction into a live scene for discussion with a team or client. >
• Brand Experience: explore interactive 3D showcases, catalogs and explanatory demos where the user acts directly with the object. > Learning to design simulations that adjust the next step to the actions of a person, rather than repeating one video for everyone.
</ >>

The practical principle for the team is not to start with “build the whole site through the model”, but with one scene without critical data. It needs to be checked for latency, readability, preservation of brand rules and clear access to the usual interface. If these conditions are not met, the generative scene remains a spectacular experiment, not a product function. </ p. >>

What does this change for design?

>

Solaris raises a useful question: where is the predetermined scenario more important for the product, and where is the sense of direct interaction? The designer will increasingly have to design not only screens, but also the boundaries of the behavior of the model: what objects it can change, what data to rely on, when it must show normal control and how not to lose availability. This is a more realistic next step than the promise of the end of front-end development. </ p. >>

Status: Runway invites partners to Early Access. There is no public mass release and evidence of suitability for combat products. >

Source: Runway - Introducing Solaris, August 31, 2026. >