Independent guide · Research preview

Runway Solaris: A Complete Guide to the Interface World Model

Runway Solaris is an early-access Interface World Model that generates a visual interface continuously, frame by frame, as a person clicks, drags or gives instructions. This guide explains what Solaris is, how it differs from Gen-4.5, what it may be useful for, its current limitations and how to request access.

By Runway Aleph

Concept illustration of a visual interface reshaping around a pointer interaction in real time
Original concept illustration for this independent Solaris guide; it is not an official Runway product screenshot.

Runway Solaris at a glance

Model category

Runway calls Solaris its first Interface World Model: a model that renders the interface and its responses together.

Output behavior

It synthesizes the next visual frame continuously in response to clicks, drags and other inputs instead of returning a fixed video clip.

Technical lineage

Solaris builds on Runway Gen-4.5, adapted to understand interaction and produce frames quickly enough for real-time use.

Availability

As of September 1, 2026, access is limited to an early-access request form; public pricing and API access have not been announced.

Definition and release

What Is Runway Solaris?

Runway introduced Solaris on August 31, 2026 as the first model in a new family it calls Interface World Models. The simplest definition is: Solaris generates an interface as you use it. Rather than compiling a visual design into HTML, CSS, JavaScript or another intermediate representation, the model synthesizes the visible application frame and predicts how that frame should change after an input.

That makes Solaris different from both a static UI generator and a conventional no-code builder. A static generator can draw a screen, while a no-code builder still produces a structured application whose behaviors were defined in advance. Solaris treats the image itself as the interactive environment. A click, drag or instruction conditions the next frame, so the response can be visual, continuous and dependent on the session state.

It is also not a general public replacement for coded websites today. Runway presents Solaris as an early research step and is working with selected partners before a public launch. The announcement demonstrates a direction for generated software, not a production guarantee that every interface, accessibility requirement or business workflow can already be handled without code.

FactCurrent Solaris status
AnnouncementAugust 31, 2026
DeveloperRunway
CategoryInterface World Model
Core behaviorContinuous frame generation conditioned on interaction
Visual targetQuality that holds at 720p
AccessPartner early-access request
Public price or APINot announced as of September 1, 2026

Availability and commercial terms can change. The date above is the last verified date for this guide, not a promise of future access.

What is genuinely new

Solaris Key Features: Visual, Alive and Open-Ended

Runway groups Solaris around three ideas: the interface is entirely visual, it remains alive while the session continues, and it is open-ended. These labels matter because they describe a different software abstraction. The model is not only decorating a predefined component tree; it is responsible for the visible state and for the transition caused by an action.

Entirely visual means a starting image can become the application surface. Runway's examples include directly manipulating products in a virtual store rather than selecting them through menus. Alive means the scene can keep evolving between explicit commands: reflections, objects and ambient motion may continue to change while the user interacts. Open-ended means the response does not have to be one of a small set of event handlers anticipated by a developer.

The practical opportunity is a prompt-to-interface workflow in which a visual concept becomes interactive without first being reduced to a rigid representation. The practical risk is equally important: deterministic software is easy to test against known states, while a generated interface can produce a plausible but incorrect state. Solaris therefore looks most credible first in exploratory, visual and low-risk experiences, with verified data and conventional software surrounding it where accuracy matters.

Real-time interaction

Clicks, drags and instructions become conditioning signals for the next generated frames, allowing the scene to respond continuously.

Session coherence

The model is designed to preserve layout, objects and meaning across a session rather than only within one short rendered clip.

Direct visual fidelity

By operating on the visual interface, Solaris aims to avoid information lost when a detailed design is translated into a simpler coded representation.

Unscripted behaviors

The same starting scene can respond to different requests without every possible transition being implemented as a predefined workflow.

Model behavior and applications

How Solaris Works and What You Could Use It For

Solaris builds on Gen-4.5, Runway's video generation model. Standard video diffusion systems can spend many refinement steps producing a complete clip, which is too slow for direct interaction. Solaris instead generates sequentially and uses only interactions that have already happened to condition what comes next. In effect, each new frame depends on the visible state and the user's recent action.

Runway says interactive systems begin to feel unresponsive at roughly half a second of delay, so speed, coherence and serving cost are linked engineering constraints. The announcement says the work that made Solaris real time also made it substantially cheaper to run than a standard video diffusion model, but it does not publish an end-user price. Do not interpret lower model cost as a free or inexpensive public product until commercial terms exist.

The clearest near-term uses are highly visual interactions: an adaptive storefront, a spatial product configurator, an interactive tutorial, a simulation that reacts to direct manipulation, or an exploratory prototype that would be expensive to hand-code. Runway also highlights agent training. A generated interface can vary its layout continuously, which may help computer-use agents learn concepts that transfer beyond one fixed website.

  • Interactive education and demonstrations that render the next step in context
  • Visual commerce experiences grounded in real product imagery and verified data
  • Rapid experience prototypes where the interaction space is still being explored
  • Adaptive tutorials that recover when a user deviates from a fixed sequence
  • Dynamic training environments for agents that must handle changing interfaces
  • Creative installations and spatial interfaces where continuous visual response is the product

Model comparison

Runway Solaris vs Gen-4.5: Interface Model or Video Model?

Solaris and Gen-4.5 share technical lineage, but they solve different jobs. Gen-4.5 accepts text or an image and returns a finished video lasting 2–10 seconds. Solaris begins from an interface state and keeps generating the next frame while the user interacts. One creates media to watch; the other is intended to create an environment to operate.

The distinction changes how quality is measured. For Gen-4.5, prompt adherence, motion, visual fidelity and the finished clip are central. For Solaris, latency, session coherence, legible interface state, correct responses, accessibility and integration become just as important. A beautiful two-second interaction is not enough if text drifts or a long session loses the meaning of an object.

Gen-4.5 is already available on Runway's web product and developer platform with published credit pricing. Solaris is a partner research preview without public pricing or a documented public API. If your goal is to make a video today, use a video model. If your goal is to explore a continuously generated visual application, Solaris is the relevant research direction.

DimensionSolarisGen-4.5
Model typeInterface World ModelVideo generation model
Primary outputContinuous interactive visual stateFinished video clip
User inputClicks, drags, instructions and starting contextText or text plus image
DurationSession-based; public limits not announced2–10 seconds per generation
ResolutionDesigned to hold visual quality at 720p720p base output
AccessEarly-access requestWeb and Runway Dev
Published pricingNone yet12 credits per generated second
Best fitGenerated apps, simulations and agent environmentsText-to-video and image-to-video shots

Gen-4.5 specifications and prices are current as of September 1, 2026 and may change.

Read the complete Runway Gen-4.5 guide

Research-stage boundaries

What Solaris Cannot Reliably Do Yet

Runway identifies four major limitations. First, stable and legible text remains difficult, even though interfaces depend heavily on text. Second, generated output can look convincing without being correct, which creates a trust problem for instruction, commerce and other consequential tasks. Third, visual and semantic coherence can degrade over long, open-ended sessions. Fourth, a generated visual surface still has to connect to screen readers, accessibility APIs and the rest of a production software stack.

These are not edge cases; they define whether Solaris can move from a compelling demo to dependable software. Text errors can make a control unusable. An incorrect product or instructional response can harm a user. State drift can invalidate a multi-step task, and a purely visual surface can exclude people who rely on assistive technology.

A responsible evaluation should test failures, not only showcase moments. Use verified reference material, compare repeated runs, inspect text and object identity over time, test recovery after unexpected actions, and include keyboard and screen-reader requirements from the beginning. Keep critical transactions and factual claims anchored to deterministic systems until the generated layer can be validated.

Text stability

Generated video still struggles to keep interface text consistently readable and unchanged in real time.

Trust and grounding

A visually persuasive answer may still be wrong; verified images, documents and product data are needed for grounding.

Long-session drift

Objects, layout and meaning can lose coherence as an open-ended interaction continues.

Accessibility and integration

Generated pixels must still work with assistive technologies, APIs, data systems and conventional application controls.

Availability

How to Access Runway Solaris

As of September 1, 2026, Solaris is not listed as a self-serve model in Runway's normal tool selector or public developer model catalog. Runway says it is working with key partners toward a public launch and provides an early-access form at the end of the official Solaris announcement.

The form asks for a full name, business email, company name, company size and a short description. Use the description to explain the experience you want to build, why a continuously generated interface is necessary, what verified data or assets will ground it, and how you will evaluate latency, coherence, safety and accessibility. Runway has not published selection criteria, approval timing or pricing, so avoid claims that any particular company profile guarantees access.

Before applying, decide whether you actually need an interface world model. If a coded prototype, a Gen-4.5 video or a small set of deterministic interactions can validate the idea, start there. An early research preview is most useful when the open-ended visual response itself is essential to the product or experiment.

  • Read the official announcement and review the demonstrated interaction patterns.
  • Write a concrete use case with users, inputs, expected responses and failure boundaries.
  • Identify the product images, documents or structured data that would ground the experience.
  • Define how you will test latency, session coherence, correctness and accessibility.
  • Submit the early-access form with accurate business and project information.
  • Treat access, pricing, API availability and launch timing as unconfirmed until Runway responds.

Direct answers

Runway Solaris Frequently Asked Questions

Continue exploring

Place Solaris in the Runway model family

Solaris is easier to understand beside the video model it builds on and the broader Runway model directory. These links stay informational and do not imply that Solaris is available in our video generator.