When you’re sketching out an app idea on the back of a napkin, there’s a long road between that napkin and a working interface. Designers painstakingly craft pixel-perfect mock-ups in Figma or Sketch, then hand them off to developers who translate those visuals into HTML, CSS, and JavaScript. It’s a workflow riddled with busywork: exporting assets, slicing images, chasing down hex codes, wrestling with responsive layouts, and squinting at spec sheets. Google’s new Labs experiment, Stitch, aims to collapse all that friction into a single generative AI-powered leap—from rough prompt or sketch to production-ready front-end code in mere minutes.

Stitch emerged from a simple insight: what if a single tool could serve both designers and developers, leveraging the same AI brain? Behind the experiment is Gemini 2.5 Pro, Google’s latest multimodal powerhouse tuned for both language and visual tasks. By feeding it plain-English instructions (“Build a gallery app with a dark theme and card layouts”) or uploading a wireframe sketch, you get back a fully fleshed-out interface complete with style guide, interactive components, and exportable code. No more exporting PNGs and manually writing boilerplate—you’re handed a working UI ready to drop into your project.

At its core, Stitch provides two entry points:

1. Natural language prompts: Simply describe the application you envision—specify color palettes, layout preferences, user interactions, even accessibility considerations. The AI interprets your description and generates a mock-up that reflects your vision.
2. Image inputs: Got a rough sketch on a whiteboard? A screenshot of an existing app you like? Stitch ingests wireframes or photos of hand-drawn layouts and transforms them into polished digital designs.

Both modes exploit Gemini 2.5 Pro’s multimodal strengths, seamlessly bridging text and vision. The result is a live preview you can iterate on instantly—no more context-switching between design and code.

Design is rarely a one-and-done exercise. Recognizing that, Stitch lets you spin out multiple variants of any interface with a single click. Want to explore different button shapes, typography scales, or grid structures? Generate a handful of options side by side, compare them, and pick the winner. This variant feature dramatically accelerates experimentation, shining a light on possibilities you might never have considered if you were hand-coding each tweak.

Once you’re happy with a design, Stitch gives you two natural pathways forward:

• Paste to Figma: The generated UI can be injected directly into your Figma project, complete with editable layers, component organization, and style tokens. Designers can then refine spacing, swap fonts, or integrate with existing design systems.
• Export front-end code: Alternatively, grab the HTML, CSS, and JavaScript that Stitch produces. It’s structured, modular, and ready to be integrated into your codebase—no glue code required.

This dual export strategy acknowledges that many teams already have established workflows in Figma, while others prefer to jump straight into code. By supporting both, Stitch positions itself as a connective tissue rather than a replacement for familiar tools.

It’s no accident that Google chose to highlight Figma integration. Figma has become the de facto hub for collaborative design, and earlier this month, the company introduced Make UI, a tool to generate basic interfaces. Stitch, however, goes further by coupling design generation with production-grade code. For teams evaluating whether to stay in the Google ecosystem or lean on Figma, Stitch offers a powerful incentive: skip the hand-off entirely and have your interface live-coded the moment you hit “Generate.”

Gemini 2.5 Pro, the model powering Stitch, represents Google’s latest push in developer-focused AI. It not only interprets complex visual layouts but also understands front-end frameworks, CSS conventions, and responsive design patterns. Earlier this I/O, Google showcased how Gemini 2.5 Pro can translate video walkthroughs into code snippets and enhance Code Assist in IDEs. Stitch is the first public experiment to marry those capabilities with pure UI generation—an ambitious testbed for what multimodal AI can do in a real-world design-dev pipeline.

Stitch is in experimental form on Google Labs, free to try for anyone curious enough to join the waitlist. It currently supports English prompts, with additional languages on the roadmap. As an experiment, it’s unlikely to replace seasoned UX teams overnight, but it will appeal to:

• Solo developers building MVPs who need quick, polished interfaces without hiring a designer
• Product teams looking to prototype dozens of variations in hours rather than days
• Design-dev hybrids seeking a more integrated workflow that minimizes context-switching

Over time, expect Google to expand Stitch’s language support, tighten Figma integration (perhaps real-time collaboration in Google AI Studio), and increase output fidelity for complex patterns like animations or stateful components.

Stitch is more than just a novelty; it’s a tangible peek at where AI could take software development. By collapsing the gap between idea and implementation, tools like this could usher in an era where designers and developers speak the same AI-native language. Whether that future belongs to Google, Figma, or another innovator remains to be seen—but for now, Stitch is a compelling argument for a world where a simple prompt can spark a fully functional UI in moments.