Runway unveils Solaris: The first Interface World Model
机器之心 · wechat · 2026-09-01
Runway introduced Solaris, defined as the first "Interface World Model." Instead of generating code for interfaces, it uses a model to directly generate the next visual frame in real-time based on user interactions (clicks, drags), making the visual itself the software.
Core Mechanics:
- Built on Gen-4.5, it conditions generation on user inputs to predict visual states post-action.
- Achieved real-time speeds via autoregressive generation and distillation of denoising steps.
- Architecture separates the LLM (reasoning/decision) from the World Model (rendering).
Experimental Results:
- Direct visual generation results in less information loss for complex interfaces compared to the "code-to-render" pipeline.
- In 7,500 human comparisons across 30 cases, Solaris significantly outperformed code-based solutions in following instructions and natural behavior.
Applications:
Beyond UI generation, Solaris aims to serve as a training environment for Computer Use Agents. By generating infinitely varying interface layouts, it forces agents to learn general interaction logic rather than memorizing specific button positions.
Related event: Runway Unveils Solaris, Its First Interface World Model(16 posts)→
More from Multimodal
- Drama-3 weights land on the Hugging Face Hub with an official demo — Gradio · 2026-10-16
- MiniCPM-V 4.5 offer open-source local image understanding to replace paid vision APIs — heyshrutimishra · 2026-10-16
- Surreal AI-generated landscape hides "another landscape inside the measurements" — piotrbinkowski · 2026-10-16
- techhalla: Building 3D scenes in Blender is the key to AI filmmaking — techhalla · 2026-10-16
- Fan animation made in Blender with Claude Code (Opus 5.5) — Icy-Appointment3583 · 2026-10-16
- Making Blender animations with Opus, Astra and Argon: how you prompt still matters — chongdashu · 2026-10-16