A single-page gallery of all 45 images from OpenAI's ChatGPT Images 2.0 launch, with the original prompts and a direct "Try in ChatGPT" link for each.
Single index.html, no build step. All 45 prompts are embedded directly in JS. No fetch, no CORS, no server — just open the file or push to GitHub Pages.
Dark mode default. Warm near-black (#0e0d0a) so images read as the primary content, not the chrome.
Two views, one toggle. Grid is for visual browsing; Scan puts a thumbnail on the left and the full prompt text in the remaining space so you can skim all 45 prompts without opening anything.
Filter by category. Six categories derived by looking at every image and prompt, not guessing:
- Photography (15) · Comics & Manga (9) · Poster & Ad (10) · Infographic (5) · Editorial (3) · Illustration (3)
Lightbox with ?id=N deep links. Click any image to open a full-size lightbox. The URL updates so individual images are shareable. Arrow keys, j/k, and Escape work. Filters carry into the lightbox — navigation skips filtered-out items.
"Try in ChatGPT" via ?q= Opens ChatGPT with the prompt pre-filled. No known URL param forces image generation mode; ChatGPT recognizes image prompts automatically.
Fonts: Sora (UI) + Source Serif 4 (prompt text) via Google Fonts.
All images are AI-generated outputs collected from OpenAI's public announcement. Prompts sourced from chatgpt-images-2-0-prompts.md, which mirrors the original blog post credits.
index.htmlis the whole app: structure, styles, image data, and lightbox logic all live there.- The lightbox contract is stable:
#lightbox,#lb-prev,#lb-next,#lb-close,#lb-copy-btn, and#lb-chatgpt-linkare the primary hooks. - Opening a lightbox image pushes
?id=N; closing it returns to the base pathname. - Keyboard controls while the lightbox is open:
ArrowLeft/kfor previous,ArrowRight/jfor next,Escapeto close. - Local-only harness docs can mirror the selectors and URL contract above without needing changes to the shipped site.