Runway's Solaris generates app interfaces in real time
The system renders each frame of the interface at 720p instead of running code, reacting to clicks, drags and voice. No launch date yet.
Illustration: a translucent glass surface with fine lines of light running across it, as if the surface were being redrawn frame by frame.
Solaris is a Runway research model that does not build an interface out of code but generates it as video at 720p, one frame at a time, while you use it.
At a glance
- Runway calls Solaris the first model in a category it labels "Interface World Models".
- The interface is rendered frame by frame at 720p rather than executed as code.
- It is built on Runway's Gen-4.5 video model, continuing the line started with GWM-1.
- A language model decides what changes; the world model draws the resulting frames.
- Status: research only. Early access via a form, with no date and no pricing given.
Solaris, unveiled by Runway, does not assemble an application's interface from program code. It generates it — every frame at 720p, produced as you click, drag or speak. Runway calls it the first model in a category it labels Interface World Models.
How it works
Conventional software goes from design to code to a running app. Solaris drops the middle step. A language model decides what should change on screen, and the world model renders the frames that follow.
The foundation is Runway's Gen-4.5 video model, extending the direction the company opened with GWM-1. The interface is, at bottom, moving footage rather than a layout a browser paints from markup.
The use cases Runway proposes
Three areas are named. In e-commerce, the store itself becomes the interface: you drag a shirt onto a picture of yourself to try it on, and the presentation adapts to the individual shopper instead of following one fixed template.
For product visualization, spoken commands such as "move the table" or "change the couch color" produce an immediate visual response — aimed at furniture and interior design work. The third area is learning: rather than answering in text the way a chatbot does, Solaris shows the next step visually inside the relevant context.
The bigger claim
The demos matter less than the argument behind them. Runway suggests the app as a fixed unit could stop making sense. Getting anything done today means opening the right prebuilt application — one for shopping, one for news, one for reservations.
If an operating system can generate a fitting interface on demand, the reason to sort software into a fixed catalog disappears. Runway also argues that generated interfaces could help AI agents, which currently stumble over layout variations and unconventional websites.
What does not work yet
Runway is direct about the gaps. Keeping text stable and legible remains, in its own description, a hard problem — in a medium where interfaces depend almost entirely on text.
Reliability across longer sessions is unresolved, as is the risk that output looks convincing while being wrong. Accessibility is the sharpest open question: a screen reader cannot read a rendered video frame, which leaves support for assistive technology unsettled.
Reading the announcement
Solaris is research, not a product. Runway says it is looking for launch partners and is granting early access through an application form; the announcement gives no date, no price, and no figures for latency or running cost.
Runway also leaves the central question open on purpose — whether people will accept an interface that looks different every time they open it. Predictability has so far been an asset of fixed software, not a defect.
FAQ
What is Runway Solaris?
A Runway research model that generates user interfaces as video frames at 720p instead of building them from executed code. Runway calls the category "Interface World Models".
Can you use Solaris today?
No. It remains a research effort; Runway is seeking partners and offering early access through a form. The company has not given a launch date or a price.
Why is text a problem for generated interfaces?
Because the system paints each frame anew rather than setting type. Runway itself calls stable, legible text rendering a hard problem, and lists missing screen reader support as an open issue.
