The UI Prompt Generator turns a screen type, platform, design style and industry into three ready-to-paste AI design prompts: a short one, a full structured brief, and one written for code tools. It runs entirely in your browser, free and unlimited. Built by GenDesigns, which turns prompts like these into real screens.
Picking one swaps in realistic data — balances, workout names, product rows — instead of placeholder text.
Separate with commas. Recognised words like “accessible” or “dark mode” also pull in the specific rule they imply.
One paragraph for a chat box. Everything essential, nothing a model will skim past.
Design a home dashboard for Northwind, a consumer app. iPhone 16, 393 × 852 pt, portrait. Apple Human Interface Guidelines, native iOS controls. It is opened by someone who opens the app two or three times a week, one-handed, with about thirty seconds of attention. Show a hero figure reading 128 labelled "This week", a 7-day bar chart of items completed, and four item rows: Quarterly review — updated 2h ago. Style: Modern minimal — white canvas, one accent colour, hairline borders instead of shadows, a lot of empty space doing the work. Use bottom tab bar with 4 tabs (Home, Library, Activity, Profile). Output HTML and Tailwind CSS, minimum touch target 44 × 44 pt, and use the exact numbers and copy above instead of placeholder text.
The link carries every setting, so whoever opens it sees the same three prompts.
A prompt is the input. GenDesigns is what turns it into real, themed app screens you can open, edit and export as HTML and Tailwind. Plans start at $10 a month, credits roll over and never expire.
Generate these screens“Design a clean, modern fitness dashboard” is a sentence every app could have written. A model reading it has no platform, no user, no content and no constraints, so it fills the gaps with its safest defaults: blue on white, a generic sans-serif, evenly weighted cards. That is why so much AI-generated UI looks like the same app.
Each field on this page removes one of those gaps. The generator composes them into the six-part PROMPT structure we documented after generating thousands of screens.
P — Platform & device
iOS, Material 3 or responsive web — each has different navigation, type and touch rules.
R — Role & user
Who opens the screen and in what moment. A dashboard for a CEO is not a dashboard for an analyst.
O — Output specification
The exact elements and real data. $1,299.00 occupies different space than $42.50, and the layout should reflect that.
M — Mood & style
A named style with real rules — surfaces, radii, depth, motion — instead of an adjective.
P — Patterns & components
Named navigation and layout patterns, so the parts belong together instead of colliding.
T — Technical constraints
Contrast ratios, touch targets, breakpoints, output format. Constraints improve output.
The three outputs differ on purpose. The concise one is a paragraph for a chat box, where a long prompt gets skimmed. The detailed brief spells out every part of the framework and adds a DO NOT block, because naming the failure modes — Frankenstein layouts, lorem ipsum, two competing primary buttons — prevents them more reliably than describing the ideal. The third is written as an engineering ticket: a design-token contract, a numbered structure, copy to render verbatim, the states to build, and acceptance criteria the output can be checked against.
The full reasoning, plus the advanced techniques this tool does not cover — visual references, the three-pass refinement method, design system anchoring — is in the complete framework for AI UI prompts. If you would rather browse finished prompts than compose one, the library of 100+ copy-paste AI design prompts is organised by screen type, style and industry.
A UI prompt is the instruction you give an AI tool to generate an interface. A weak one names a screen ("design a dashboard"); a strong one also fixes the platform, the person using it, the exact content, the visual style, the layout patterns, and the technical constraints. That specificity is the whole difference between generic output and something you would show a client.
Six things, which we call the PROMPT framework: Platform and device, Role and user, Output specification with real data, Mood and style, Patterns and components, and Technical constraints. Most weak prompts skip the last four. The generator on this page fills all six for you, and adds an explicit DO NOT block, because telling a model what to avoid is as useful as telling it what to build.
Any of them. The prompts are plain text with no tool-specific syntax, so they work in GenDesigns, v0, Cursor, Bolt, Lovable, Figma Make, ChatGPT and Claude. The third variant is written specifically for code-generating tools — it declares a design-token contract, verbatim copy, required states and acceptance criteria, which is the shape those tools follow most reliably.
Name a style that has an actual definition instead of saying "clean and modern", which every app claims and no model can act on. Glassmorphism, neobrutalism, dark premium, editorial, neumorphic and Material 3 all carry real rules about surfaces, radii, shadows and type. Then pin the specifics: a hex primary, a corner radius, a font, and one sentence on how depth is handled.
The generator composes screen-level prompts, because a screen is where layout decisions actually get made. For a single component, take the generated prompt, keep the design-token block and the technical constraints, and replace the structure section with the component and every variant and state you need — default, hover, focus, disabled, loading and error.
Yes. It runs entirely in your browser with no model call, no account and no limit on how many prompts you generate, which is also why it is instant. Generating the actual screens from those prompts is what GenDesigns does, and that is a paid product starting at $10 a month; credits never expire and roll over month to month.
