Google rolls out Gemini Sketch: multimodal model for generating Material UI components

Tech · 5 min read

Gemini Sketch can take a photo or finger-drawn sketch and output editable Material UI components, including accessible markup, elevation, and responsive constraints. Outputs are available as code snippets, XML/Compose manifests, or downloadable assets for popular design tools.

The model leverages Google's internal datasets of component patterns and accessibility rules to ensure generated components meet baseline WCAG guidance by default. It also provides a 'style pulse' control to shift between Google Material, Fluent, and iOS-styled components, which Google said is implemented through a controlled style transfer layer.

Privacy-forward deployment options, including on-prem and private-cloud hosting for enterprise customers, were announced alongside the product. Designers expressed excitement about rapid prototyping possibilities, while some called for clearer ownership and export guarantees for generated code and assets.