Google unveils 'Gemini Frame'—an LLM focused on UI component generation

AI · 5 min read

Gemini Frame is a specialized offshoot of Google's Gemini line, trained with a data mix emphasizing UI screenshots, component libraries, and front-end code. Google says the model improves on previous generations in producing accessible HTML/CSS, React components, and platform-specific layout hints for iOS and Android.

The model offers a 'screenshot-to-structure' mode where developers can upload a snapshot and receive a component tree, CSS rules, and suggested accessibility labels. There's also a 'design advisory' output that recommends interaction states and animation timings based on common patterns.

Google is releasing Frame via an API and a beta plugin for popular design tools. The company highlights enterprise governance controls and a dataset provenance dashboard to address designers' concerns about reused patterns and IP.