Runway

AI · Developer Tools · Generative Video

Launch

Runway launches Solaris, a real-time AI model that generates UIs frame by frame

The system skips coding entirely, rendering visuals and handling user interaction in one continuous generative loop.

• Runway unveiled Solaris on August 31, 2026, calling it the first model in a new category it calls Interface World Models, generating interactive interfaces frame by frame with no code.
• Runway is a New York AI lab founded in 2018 by Cristobal Valenzuela, Anastasis Germanidis and Alejandro Matamala Ortiz, originally focused on machine-learning tools for artists and video generation.
• Traditional software requires converting visual design into code before it can run; Solaris instead renders and responds to interactions jointly, eliminating that intermediate translation step.
• Solaris builds on Runway's Gen-4.5 video model, generating frames autoregressively before being distilled into a faster model for more stable, interactive-speed 720p output.
• Runway claims Solaris outperforms frontier LLMs when generating new interfaces on measures of structural similarity and information retention, though no latency figure was published.
• Solaris is a research preview, not a public product; Runway is accepting early-access requests and working with partners toward a broader launch, with no release date, API or pricing announced.
• By treating interface generation as a visual world-modeling problem rather than code generation, Runway is betting that AI agents will navigate software more reliably if the interface itself adapts dynamically rather than being frozen in advance.