Google DeepMind releases Gemini Designer APIs for multimodal prototyping

AI · 5 min read

Gemini Designer provides endpoints that accept hand-drawn wireframes, screenshots, or text prompts and return structured prototypes including component placements, CSS/SwiftUI snippets, and suggested user flows. The system emphasizes end-to-end prototyping from rough sketches to interactive mockups.

Google focused on developer ergonomics: the API returns standardized JSON schemas for component trees and accessibility annotations, making it straightforward to map responses into various front-end frameworks. SDKs for web and mobile ship with starter templates.

DeepMind also released a playground for designers to experiment with sketch-to-prototype conversions and to export assets to Figma, React, and Flutter. The demo highlighted iterative prompting where designers can refine layouts by adjusting high-level constraints.

Privacy and evaluation were noted: Google said the models underwent in-house audits for hallucination in code-generation and that the team is building fine-grained controls for enterprise clients to scope generation behaviors.