Galileo AI

Galileo AI was an AI-powered UI design startup founded in 2022 by Arnaud Benard and Helen Zhou that generated editable, high-fidelity interface designs from natural language descriptions. It was acquired by Google in May 2025 and relaunched as Google Stitch, powered by Gemini models. Stitch has since evolved into a full AI-native software design canvas available for free through Google Labs, designed for product designers, UX researchers, product managers, and solo founders who want to rapidly produce visual prototypes without starting from a blank canvas.

History and Evolution

Galileo AI launched in 2022 as one of the first tools capable of turning text prompts into polished UI designs, quickly building a following among product teams. Google acquired the startup in May 2025 and immediately relaunched it as Stitch at Google I/O 2025, integrating Gemini models and deprecating the standalone Galileo product with a 30-day migration window. In December 2025, Gemini 3 integration improved layout quality and introduced interactive prototypes. The major Stitch 2.0 update in March 2026 brought an infinite canvas, voice input, multi-screen generation, DESIGN.md, and an MCP server with SDK. At I/O 2026 in May, Google added a real-time streaming design agent and simultaneous multi-user collaboration, directly challenging Figma.

How It Works

Users describe the desired interface in natural language, upload a screenshot or hand-drawn sketch, paste a URL to extract design tokens, or speak directly to the canvas using voice commands. Stitch generates up to five interconnected, high-fidelity screens within 30 to 60 seconds using Gemini 3.0 models. The AI-native infinite canvas displays all screens side by side, and a Design Agent tracks project history and reasons across the entire context. Users can refine specific elements through local prompt windows, generate variations with adjustable creative range, or make direct edits to text and images on the canvas. The Agent Manager enables parallel exploration of multiple design directions simultaneously.

Key Features

• Text-to-UI generation producing complete, multi-screen interface designs from plain language prompts, generating up to five coherent screens per prompt with consistent design systems
• Image-to-UI conversion that transforms uploaded screenshots, hand-drawn sketches, or competitor UI screens into polished, editable layouts
• URL extraction that fetches any public webpage, extracts design tokens such as colors, fonts, and spacing, and generates screens matching that visual style
• Voice Canvas supporting real-time voice commands for design changes, critiques, and interactive interviews where the agent asks clarifying questions before generating
• Figma export with clean Auto Layout, semantic layer names, and proper component grouping that preserves editability
• Code export generating semantic HTML with Tailwind CSS, React components, and integration with Google AI Studio, Claude Code, and Antigravity for full-stack development
• DESIGN.md, an agent-readable markdown file that encodes design system rules including colors, typography, spacing, and component patterns, portable across projects and tools
• MCP server and public SDK that allow AI coding agents and custom tools to query and interact with Stitch projects programmatically
• Real-time streaming design agent that shows generation progress live on the canvas and allows mid-generation steering
• Multi-user collaboration enabling simultaneous editing on the same canvas, added at I/O 2026
• Instant Prototyping that connects screens into clickable, interactive flows with automatic navigation mapping
• Variations panel generating multiple design alternatives with adjustable creative range from Refine to Reimagine
• AI-generated heat maps predicting user attention and click patterns on generated interfaces

Use Cases

• Product designers rapidly exploring layout directions and visual concepts during discovery and ideation
• Founders and product managers creating visual mockups for stakeholder presentations, investor decks, and pitch materials
• UX researchers generating screen variations for user testing sessions and comparative design evaluation
• Development teams receiving design handoff through structured Figma files, MCP-connected coding agents, and exportable code
• Non-designers and solo builders who need to communicate design intent without Figma expertise

Pricing and Access

Google Stitch is available for free through Google Labs. Users receive 350 standard generations per month using Gemini 3.0 Flash and 200 experimental generations per month using Gemini 3.0 Pro, totaling 550 generations at no cost. No credit card is required. The legacy Galileo AI paid tiers, ranging from $19 to $39 per month, have been retired following the acquisition. Google has not announced long-term pricing, and Stitch remains a Labs experiment subject to change or discontinuation.

Limitations

• Output is design-first and not production-ready code; generated interfaces require manual refinement, with roughly 60 to 70 percent of screens needing adjustments for production use
• DESIGN.md supports basic design tokens but lacks the depth of full component libraries found in Figma or established design systems
• Google Labs beta terms may permit input data use for model improvement, which may not suit teams with contractual data restrictions
• No version history, branching, or team workspace management beyond basic snapshots
• Hard cap of 550 generations per month with no overflow or rollover
• An experimental mode limitation: image input and Figma export cannot be combined in a single workflow