Stitch 3.0: AI UI generation with direct Figma export

What it is. Stitch is Google’s AI UI design tool. You describe a screen in plain text, and it generates a working layout for mobile or web on a live canvas. Version 3.0 adds streaming edits, so changes appear as they render rather than after a long wait, and in-place AI edits let you click into any element and prompt changes directly on the canvas. From there, one-click export sends your work to Figma, Netlify, Lovable, or Bolt.

Why it matters. The gap between “I have an idea” and “I have something to show” is where a lot of creative momentum dies. Stitch compresses that gap considerably. You are describing what you want rather than building it component by component, and you end up with something Figma-ready rather than a screenshot you then have to reconstruct by hand.

How to use it. Start with a prompt for the screen you actually need, something specific like “a mobile onboarding screen for a photo app, minimal, dark background, three steps.” Stitch renders the layout. From there, click into any section and type a follow-up prompt to change it in place, swap the typography, rework a card, adjust the hierarchy. When the screen is close enough to work from, export it to Figma and finish the details there the way you normally would. It fits naturally as a starting-point layer before your real design work begins, rather than a replacement for it.

Limits. Generated UI tends to need real design judgment applied afterward. Component naming, spacing systems, and responsive behavior will not come out production-ready, so plan for cleanup time in Figma. It works best when your prompt is precise. Vague inputs produce generic results, and the more specific your product’s visual language, the harder it is to get Stitch close without a lot of back-and-forth. It is also a Google Labs product, which historically means availability and feature continuity can shift.