Runway Solaris: A New Era of AI-Generated Interfaces

Runway has introduced Runway Solaris, the first model in a new family of AI systems called Interface World Models. Solaris is designed to generate interactive interfaces in real time, allowing websites and applications to respond continuously as users click, drag, and interact with them.

The technology explores a different approach to software design by combining visual generation with real-time interaction. Instead of relying entirely on fixed screens and predefined workflows, Solaris can generate the interface itself frame by frame as the user interacts with it.

Key Takeaway

• Runway Solaris is Runway’s first Interface World Model for generating interactive interfaces in real time.
• Instead of relying entirely on predefined screens and traditional code, Solaris generates interface frames as users interact with them.
• The system is built on Runway’s Gen-4.5 video model and its earlier GWM-1 world-model research.
• Solaris could open new possibilities for AI software development, app experiences, tutorials, shopping environments, and AI-agent training. 
• The technology is still experimental, with challenges including text rendering, trust, long-session consistency, accessibility, and cost.

What Is Runway Solaris?

Runway has launched a new AI system called an Interface World Model, called Runway Solaris. Solaris is different from a traditional website or app because it generates the interface itself, frame by frame, as the user interacts with it.

The idea connects generative AI with interactive software. Instead of translating a visual interface into an intermediate representation such as code, Solaris generates the interface directly as users interact with it.

This could be a big step forward for AI software development, where developers are looking for ways to create software that can adapt to the user.

How Does Runway Solaris Work?

Runway Solaris treats user actions like clicks and drags as inputs that influence the next frame. Solaris doesn’t just follow predefined screens. Instead, it uses these interactions to work out how the visual environment should respond and evolve in real time.

Runway says Solaris was adapted from its Gen-4.5 video-generation model and builds on its GWM-1 world-model research. It uses an autoregressive approach, where each frame depends on what came before, while a reduced generation process helps make the system fast enough for interactive experiences.

How Does Solaris Combine Reasoning and Rendering?

Solaris separates deciding what should happen from showing how it happens on screen. A language model understands the user’s request, determines how the scene or interaction should change, and creates instructions for Solaris.

Solaris then generates the interface frame by frame, using the user’s clicks, drags, and typing to shape what happens next. The language model decides what the experience should do, while Solaris determines how it should look and respond in real time.

Why Is Solaris Different From Traditional Software?

Traditional applications use code, UI components and predefined workflows. This makes them predictable and easier to test and integrate. But developers have to plan for how users will interact with the software, which can make it less flexible.

It is different because it creates the interface as users interact with it. This combines how things look and how they work in real time, creating new possibilities for designing AI interfaces that go beyond fixed layouts and predefined components.

Solaris AI Makes Interfaces More Visual

One of the most important ideas behind Solaris AI is that interfaces do not have to behave like normal websites. Instead of using pages, users can interact directly with objects and scenes in a more natural way.

For example, imagine a virtual shop where you can pick up a shirt and drag it onto a picture of yourself to try it on. Runway’s examples show how these types of spaces can respond to user actions more like physical spaces than traditional web pages.

This could make AI-powered interfaces more visual, flexible, and responsive to the user’s context. Instead of just clicking buttons and menus, users could interact with the interface itself to complete different tasks.

From Fixed Pages to Interactive Environments

Traditional websites are usually organised into pages.

You have:

• Home page
• Product page
• Checkout page
• Settings page
• Confirmation page

Solaris has something new.

The interface can be more like a world that users interact with directly. For example, Runway lets you move objects, change colours, mix ingredients or explore different visual environments by interacting with them directly. This could make an AI user interface feel less like navigating traditional menus and more like interacting with a dynamic digital world.

What Could It Mean for AI Software Development?

The effect on AI software development could be big, but it’s too early to say that Solaris will replace normal development. Today, developers convert designs into code and connect interfaces with databases, APIs, business logic, authentication, and other systems. Solaris looks at whether parts of this visual and interactive layer can be created dynamically.

This could eventually help developers try out interfaces that are difficult or expensive to build manually. But traditional engineering will still be important for reliability, security, data handling, accessibility, and integration, especially when building software that is ready for production.

AI for App Development Could Become More Flexible

The idea behind Solaris is also relevant for AI in app development. In this area, interfaces could become more adaptive instead of relying entirely on fixed screens and workflows.

Instead of building every possible interaction beforehand, an AI system could generate parts of an experience based on what the user wants to accomplish. For example, a travel interface could adapt when someone wants to compare hotels, explore destinations, or plan an itinerary.

But this is still something that researchers are looking into. Solaris can’t yet build apps for every possible use.

What About the AI User Interface?

The idea of an AI user interface is changing as AI assistants go beyond just giving text-based answers. Solaris lets you interact with a generated environment, instead of just reading instructions or answers.

Runway shows this with an interactive combustion experience, where users can experiment with different materials and see what happens visually. This suggests that in the future, AI systems will combine conversation, visual creation, and real-time interaction into one experience.

How Could Solaris Help AI Agents?

Another important thing that can be done with it is to train AI agents. Current computer-use agents often rely on buttons, menus, fields, and predefined layouts, so even small changes to the interface can make tasks harder. Dynamic environments could help agents interact with interfaces that are always changing and develop more adaptable computer-use skills.

This could help agents develop their skills in interacting with customers, instead of simply memorising where specific interface elements are located. By trying out different layouts and behaviours, agents can learn to adapt to new digital environments.

Researchers are also looking at different ways to train GUI agents in the rest of the world, which shows that this is an active research area outside of Solaris itself. This makes it an interesting area to look at for the future of AI agents.

How Does Interface Translation Affect Software?

Traditional software translates designs into code before making them interactive. This can sometimes cause details about layout, objects, materials and appearance to be lost. They tested this issue on 30 different websites and apps, using something called SSIM to compare how similar things looked, and something called DINOv3 features to measure whether important content stayed recognizable.

The results showed that it became harder to rebuild as the interfaces grew more visually complex, especially with natural images. Solaris works differently by working directly with the visual interface while generating each new frame. This helps to keep more visual and semantic information without relying entirely on traditional translation.

Solaris and the Future of AI Interface Design

The biggest opportunity for AI interface design may be personalisation.Runway envisions a future where online stores can preserve their brand identity while adapting layouts, products, colors, materials, and recommendations to individual users. Instead of showing everyone the same website, companies could create experiences that respond to different user needs.

But this approach also brings challenges. Brands would need to keep their image the same, while making sure that the interfaces they create are accurate, trustworthy, easy to use, and useful. If we can deal with these challenges, we could see a big change in how businesses design their websites and digital experiences.

What Are the Limitations of Runway Solaris?

Although it has a lot of potential, the technology is still a very early research system. 

Runway identifies several challenges.

Text Generation: Generative video systems still find it hard to produce readable, stable text. This is important because websites and apps need accurate text.

Trust: Even if the interface looks good, it can still give the wrong information. When it comes to business or teaching, it’s crucial to be reliable.

Long Sessions: It is still hard to keep the same visual appearance and meaning throughout long interactions.

Accessibility: Interfaces that are generated will need to work with technologies such as screen readers and accessibility APIs.

Cost: Making each frame of the animation using a computer process takes more time than using a pre-built webpage. Runway says its real-time approach is much cheaper than standard video diffusion, but cost is still important.

What Makes Solaris Important?

Solaris AI is more than just a way to make screens look good. The main idea is that the interface itself could become generative, changing the traditional relationship between software and users and creating more dynamic, adaptive digital experiences.

Instead of:

Traditional model: Developers build predefined software, and users operate it.

Solaris vision: Users interact with a system that can generate and adapt the interface around their needs.

People use the technology and the technology responds and adapts. This is one reason why developers and product teams are finding AI for app development and generative interfaces interesting.

How Did Runway Evaluate Solaris?

Runway checked Runway Solaris against a coded interface using Claude Opus 5 as the system based on the language model. Both approaches started with the same image and received the same interaction requests.

The study looked at 30 different examples of interactions and asked 250 people to make nearly 7,500 judgments. Participants compared how well each result followed the instructions and how naturally it fit into the scene.

Runway said that most people preferred Solaris for 61% of the comparisons, compared to 24% for the coded result. Solaris was the best at natural behaviour, with 71%, compared with 21% for the others. These findings come from Runway’s own evaluation, so they should not be treated as an independent industry benchmark.

Runway Solaris vs Traditional Interfaces

Updates → Developer-driven → Potentially generated in real time

Feature	Traditional Interface	Solaris Approach
Interface	Pre-built	Generated dynamically
Interaction	Predefined	More open-ended
Visual rendering	Usually code-based	AI-generated
User experience	Fixed workflows	Adaptive environments
Interface changes	Developer-driven	Can respond dynamically to interaction
AI interaction	Often separate from UI	Integrated with the interface

The comparison does not mean traditional interfaces are obsolete. 

Traditional software is better suited to situations where predictable behaviour, accessibility, performance and reliability are essential. Solaris explores what becomes possible by approaching these requirements through a more flexible, AI-generated interface.

The Future of AI-Generated Interfaces

This technology is an early step towards creating digital experiences that are always changing. If it becomes faster, more reliable, controllable, and better at handling text and accessibility, it could influence websites, apps, education, digital commerce, and AI agents.

When it comes to making software that uses AI, the big question isn’t whether the interfaces that are created will replace the traditional code. The real question is how these two things can work together. A hybrid approach could combine the reliability of traditional software with the flexibility of generative AI.

Final Thoughts

Runway Solaris offers a new way to think about software interfaces, turning fixed pages into interactive environments that adapt as users interact with them. The technology is still in its early stages of development before it is released to the public, but it shows how AI could make digital experiences more flexible and personalised.

At the moment, Solaris is more of a research project than something that can replace traditional applications. Its most exciting feature is that in the future, software may not only use pre-built code, but could also understand what the user wants and create the right interface and experience straight away.

FAQs

1. What is Runway Solaris?

Runway Solaris is Runway’s first Interface World Model, designed to generate interactive interfaces frame by frame in real time.

2. What is Solaris AI used for?

Solaris AI is designed for interactive visual experiences, including generated interfaces, dynamic environments, and potential AI-agent training applications.

3. How Does Solaris Work? 

It uses user interactions such as clicks and drags as conditioning for subsequent frames, allowing the interface to respond continuously.

4. Can Solaris replace traditional app development?

Not currently. Solaris is an early research system and still faces challenges involving text, trust, accessibility, consistency, and cost.

5. What is an Interface World Model?

An Interface World Model combines understanding of user intent with continuous generation of an interactive visual environment. Runway describes Solaris as its first system in this category.