A focused library of GPT Image 2 prompts with a clear, editorial interface.
16,000+ upstream image & video prompts · 2,211 merged cases in this snapshot · 13 categories · 1-click copy
- 🎨 Neo-Brutalism Design Language — Sharp borders, hard drop shadows, monospace typography, and a raw editorial aesthetic.
- ⚡ Zero Build, Zero Framework, Blazing Fast — Built with vanilla HTML5, modern CSS custom properties, and native ES6+ JavaScript. Loads in milliseconds.
- 🗂 13 Curated Categories — UI & Interfaces, Photography, Infographics, Posters, Products, Characters, Documents, Scenes, History, and more.
- 🧭 Three Connected Libraries — The grid merges the original GPT Image 2 cases, Virloom metadata, and CC0 Midjourney V7 prompt text; source links and rights notes stay attached to the data.
- 🔍 Instant Search & Shareable Filters — Debounced keyword search across titles, tags, models, and aspect ratios; category, sort, search, and open cases are reflected in the URL.
- 📋 Structured Prompt Inspector & 1-Click Copy — Deep-dive modal view with formatted JSON/natural language prompts, parameter chips, and native clipboard integration.
- 📱 Fully Responsive Layout — Seamless user experience from 4K ultrawide monitors to mobile viewports, with an accessible navigation menu and keyboard-roving case cards.
- 🖼 Resilient Image Loading — Upstream images use a GitHub CDN fallback and an explicit unavailable state when neither source responds.
Real-time search bar, category pill selector with dynamic counts, sorting options, and responsive prompt card grid.
Click any prompt card to view model parameters, aspect ratios, source attribution, full prompt text, and copy directly to clipboard.
Source-flagged and latest cases surfaced for quick discovery; no unverified view-count claim.
lab-prompt-gallery/
├── index.html # Semantic HTML5 single-page structure
├── style.css # Neo-Brutalism design system & tokens
├── app.js # Vanilla JS state manager, filter engine & modal
├── licenses.html # Source licenses and attribution notes
├── privacy.html # Bilingual privacy notes
├── terms.html # Bilingual usage terms
├── legal.js # Shared legal-page language switch and return state
├── package.json # Root scripts for dev, verify, and capture
├── vercel.json # Vercel static routing & asset caching rules
├── robots.txt # Crawler policy and sitemap pointer
├── sitemap.xml # Public page sitemap
├── .gitignore # Git ignore rules
├── LICENSE # MIT Open Source License
├── data/
│ └── cases.json # Merged source case database & taxonomy
├── assets/
│ └── favicon.svg # Vector site mark
├── docs/
│ └── images/ # High-resolution documentation preview assets
└── scripts/
├── capture-screenshots.mjs # Automated screenshot tool via Chrome DevTools Protocol
└── verify.mjs # Data, markup, and deployment contract checks
The gallery is intentionally static. Use the root npm scripts or any local static HTTP server:
# Clone the repository
git clone https://github.com/sandbaseai/lab-prompt-gallery.git
cd lab-prompt-gallery
# Start a lightweight local server
npm run devVisit http://localhost:3000 in your browser.
npx serve .Open the directory in VS Code, right-click index.html, and select "Open with Live Server".
- Markup & Layout: Semantic HTML5, CSS Grid & Flexbox, Container-query friendly structures.
- Styling: Vanilla CSS3 with
:rootCSS Design Tokens, Neo-Brutalism hard shadows (--shadow-hard: 4px 4px 0 #1A1A1A), dark mode accents, and custom monospace typography. - Logic & State: Native ES6+ JavaScript, Fetch API, Debounced input handler, Keyboard navigation (Escape to close modal), and modern Clipboard API.
- Data Layer: Directly synced
data/cases.jsonmetadata with source-hosted image assets; CC0 Midjourney records intentionally omit images whose generation rights are not separately documented.
The synchronized grid currently contains the 541-case freestylefly/awesome-gpt-image-2 snapshot, 1,642 records from no-chili/awesome-gpt-image-2-prompts, and 28 prompt-text records from Pixmind's CC0 Midjourney V7 examples. Virloom metadata attribution is preserved, while third-party prompt/media rights remain reserved; Pixmind image files are intentionally omitted.
For broader discovery, the UI links to Toolcentral's 7,902-template GPT Image 2 catalog. Its repository is MIT-licensed, but the data bundle does not grant a separate record-level redistribution license, so it remains external until that review is complete.
Run npm run sync:upstream to refresh the three connected libraries. The merge is deterministic and assigns reserved numeric IDs to secondary records so URL deep links remain shareable.
The Submit Prompt button asks for confirmation, then opens a new GitHub Issue in the company repository. GitHub handles the final review and submission; a GitHub account is required to publish the issue.
The project includes a ready-to-use vercel.json configured for static asset caching:
Or deploy via the Vercel CLI (use the SandBase team scope when the project exists there):
npx vercel --scope sandbase- Navigate to your repository Settings -> Pages;
- Under Build and deployment, set Source to
Deploy from a branch; - Select branch
mainand folder/ (root), then click Save.
Contributions, prompt submissions, and feature ideas are welcome!
- Fork this repository.
- Add or update cases upstream, then run
npm run sync:upstreamand verify the image paths. - Commit your changes:
git commit -m "feat: add new UI design prompts" - Push to your branch and open a Pull Request.
- Code & Design: Licensed under the MIT License.
- Prompt Content & Metadata: Source-specific. Check the record's source link and license note before commercial reuse or redistribution; CC0 records in this snapshot are prompt text only.