AI Models

Runway's Solaris Generates Interfaces in Real Time, Promising the End of Fixed Apps

Runway has unveiled Solaris, an AI system that generates software interfaces live, frame by frame, in response to user interactions like clicks, drags, or voice commands. The company calls it the first 'Interface World Model' and claims the traditional app as a fixed unit will disappear. Built on Runway's Gen-4.5 video model, Solaris renders interfaces in 720p, adapting dynamically to user input. Use cases include interactive shopping, product visualization, and tutorials. However, the system is still a research effort, with challenges around text rendering, accessibility, and reliability. Runway is seeking partners for a launch and offering early access through a form.

Neura News

Neura News

Neura Market Editorial

September 1, 20264 min read
Runway's Solaris Generates Interfaces in Real Time, Promising the End of Fixed Apps

Runway has unveiled Solaris, an AI system that generates software interfaces live, frame by frame, in response to clicks, drags, or voice commands. The company calls it the first model in a new category it dubs "Interface World Models," and it claims the "app" as a fixed unit will disappear. The announcement arrived in an article published on Sep 1, 2026, written by Maximilian Schreiner for The Decoder.

A Break From Traditional Software

Traditional software runs code. The operating system decides what appears on screen, and apps stay the same until an update ships. Solaris breaks from that pattern by generating the interface itself, rather than translating design into code first. It does not run user interfaces as code; it renders them live in 720p. A language model decides how the interface changes, while the world model renders each frame. The result is a screen that adapts as the user interacts.

Runway positions Solaris as a sharp break from how software normally works. The company says an operating system can generate the right interfaces on its own, reducing the need for a fixed catalog of apps. The "app" model today means opening the right prebuilt app for each task, whether shopping, messages, or reservations. Solaris could replace those fixed apps with dynamic environments that adapt to the user.

Use Cases: Shopping, Visualization, and Tutorials

Runway lists several use cases for Solaris, starting with online shopping. In that scenario, users can drag a shirt onto a picture of themselves to try it on. For product visualization, users can say "move the table" or "change the color of the couch." For tutorials, Solaris shows the next step visually, such as a combustion demo with different materials. These examples suggest a system that turns static pages into interactive, responsive scenes.

The article analyzes that Solaris could be useful for interactive shopping or visual tutorials. It frames the system as a way to make interfaces feel alive, not just functional. The company also sees Solaris as a training ground for AI agents, which often fail at tasks like hotel bookings because they are trained on fixed layouts and cannot handle websites that deviate from them. Generated interfaces, by contrast, could adapt to whatever the agent needs.

Built on Gen-4.5, Following GWM-1

The #1 Newsletter in AI

Stay ahead of the AI curve

The most important updates, news, and content — delivered weekly.

No spam. Unsubscribe anytime.

Solaris builds on Runway's video model Gen-4.5, which handles the rendering of each frame. The system is the latest step in a path that includes GWM-1, a previous model from Runway in the same direction. The company has not said when Solaris will reach the public. For now, it is a research effort. Runway is looking for partners for a launch and offers early access through a form.

The technical approach is notable. Instead of a fixed interface rendered by a browser or native app, Solaris generates the interface on the fly. That means the screen can change with every interaction, and the system can respond to voice commands as easily as clicks. The language model decides what should change, and the world model draws it. This division of labor is what lets Solaris react in real time.

Open Questions and Hard Limits

Runway acknowledges that stable, readable text remains a hard problem to solve. Text rendering is still error-prone, and the article notes there is no support for screen readers, making Solaris a research effort for now. Long sessions and reliability are open questions as well. A wrong but convincing-looking result can be worse than none at all, the company admits.

Generated interfaces must also work with assistive tech like screen readers, Runway acknowledges. That is a significant hurdle for any system that renders interfaces visually rather than through code. The article questions whether users actually want interfaces that keep changing, and Runway leaves that question open. The company has not claimed that people will embrace dynamic screens, only that the technology can produce them.

What Comes Next

Runway's claim that the "app" as a fixed unit will disappear is bold, but the company has not shipped Solaris to users yet. The system is a research effort, and the company is seeking partners for a launch. Early access is available through a form. Whether Solaris becomes a product or stays a demonstration of what interface generation can do remains to be seen. The technology is real, but the questions around reliability, accessibility, and user appetite are still open.

Related on Neura Market

More from Neura News

AI Models

fal Opens Developer Access to Meta's Muse Image Model

fal has launched developer and enterprise access to Meta's Muse Image agentic image generation and editing model, available via the Meta Model API at $0.01 per image. Muse Image uses a planner-plus-diffuser architecture with web search, code execution, and self-refinement to improve accuracy on complex prompts. The model supports generation, editing, and reference-driven composition, and is priced to make high-volume workloads economically feasible.

Sep 1·5 min read
General

EFF Urges Courts to Reject AI Copyright Expansion, Citing History of Tech Panics

The Electronic Frontier Foundation has filed amicus briefs in two major generative AI copyright cases, urging courts to reject expanded protections based on what it calls hype and speculation. Drawing parallels to past panics over player pianos and VTRs, the EFF argues that AI tools are general purpose and capable of non-infringing uses, and that distorting copyright law would harm creativity and the public.

Sep 1·5 min read