Google Stitch is the free one. Describe a screen, or hand it an image, and it generates a UI and the front-end code to match, then pastes into Figma. That combination, free plus fast plus code-native, makes it the easiest way to get from nothing to a screen. What you do with that screen is the real question.
What Google Stitch is
Google Stitch is an AI UI design tool from Google Labs, free to use at stitch.withgoogle.com with a Google account. You describe an interface in plain language, or hand it a reference image, and it generates a UI design along with the front-end code to match. It is powered by Google's Gemini models and is the descendant of Galileo AI, a design-to-code tool Google brought in-house.
The one-line version is the fast, free way to turn a sentence into a screen. It is less a canvas you build on and more a generator you steer: you prompt, it produces, and you re-prompt to explore variations.
What it generates, and how you get it out
From a prompt like "a task manager with a sidebar, a calendar, and a dark theme," Stitch produces a full layout with components, type, color, and spacing, for web or mobile. You can hand it an image (a sketch, a screenshot, a mockup) and it will work from that instead, though it does better with a polished reference than a rough sketch.
Getting the work out is where it earns its place in a pipeline. You can copy the front-end code, which comes out as clean HTML and Tailwind CSS, or paste the design straight into Figma with editable layers and structure intact, so a designer can keep going in the tool they already use.
Where it is genuinely good
Speed and access are the whole pitch, and they are real. It removes the blank-canvas problem: a founder, a developer, or a product manager can get to a usable screen in minutes without knowing a design tool. It is free, with no card and no license, which lowers the cost of trying an idea to almost nothing.
As an ideation partner it is strong: generate a few directions, throw most away, keep one. Paired with a coding agent it becomes a front end you can iterate on rather than a picture of one.
Where it falls short
The most common complaint is that the output looks generic. Stitch reaches for familiar patterns, so screens tend to resemble each other and lack a distinct brand voice, which is exactly what you do not want when design is a differentiator. Fine-grained editing is limited: nudging a font, a color, or a position often means re-prompting rather than direct control.
Two more to plan around: generated designs frequently miss basic accessibility, such as contrast and target sizes, and need a manual pass, and the layouts are not automatically responsive, so a developer still owns the breakpoints. Treat it as a first draft, not a deliverable.
Where we'd reach for it
We reach for Stitch at the very front of a project: to explore directions cheaply, to put a real screen in front of a stakeholder in an afternoon, or to generate a scaffold a developer and a coding agent can build on. Free and code-native make it a low-risk first move.
We would not ship its output as finished design. A distinctive, accessible, on-brand interface still needs a designer, and a design system still needs a home. If you are comparing the AI design tools, the companion piece "What Is Claude Design? Anthropic's AI Design Tool, Explained" and the head-to-head "Claude Design vs. Google Stitch vs. Figma: How We Choose" cover the rest.