0toshigami/orbis-pictus
A browser where every click makes an AI redraw the whole page as a new image
Orbis Pictus is an experimental browser where there is no HTML or links, only AI-generated images shown as pages. Type anything and the app draws a page about it; click anything inside that image and it draws a brand-new page about whatever you clicked. It is an open-source homage to a site called flipbook.page.
What it does
- Typing a search query makes the app generate a single image that serves as the whole page.
- Clicking a point on that image sends a marked-up screenshot to a vision AI, which names the object under the click, and that name drives generation of a new child page.
- Since there is no click-coordinate API, the client draws a marker at the click point and sends the whole image (the 'tap trick'), reverse-engineered from the original site's network traffic.
- Content (text, layout) and visual style (material, palette, lighting) are written by two separate prompt files, so changing one style file changes the look of the entire app.
- Three layers of caching reuse earlier results for repeat clicks or identical prompts, cutting cost and wait time.
Why it matters
This project is a concrete demo of an interface built entirely from on-demand generative images instead of static markup and links, exposing both the promise and the real problems (garbled text, cost per image, caching strategy) of that idea. It's a useful reference for anyone wiring generative image and video models into an actual interactive product with prompt separation, caching layers, and page versioning.
Terms in this repo
- SSE (Server-Sent Events) · a way for a server to keep streaming events to the browser in real time
- VLM (vision-language model) · an AI model that can look at an image and describe or name what's in it
- prompt · the text instruction given to an AI model to tell it what to generate
- TAP_DEDUP · a setting controlling whether repeated clicks on the same spot reuse an earlier result instead of generating again
- page version · an edited variant of the same page, grouped together so you can switch between them
Repository description (English)
type anything, click anything inside, and it draws a new page about what you clicked. No links, no markup, every pixel made on demand. An open-source homage to flipbook.page.
Open on GitHubCoverage
Trending repos
- vorssaint/vorssaint-utilsOne free menu bar app replaces a dozen paid Mac utilities
- Alishahryar1/free-claude-codeA local proxy that lets coding AI agents run on 49 free or cheap model providers instead of one paid service
- freestylefly/awesome-gpt-image-2A library of 532 reverse-engineered prompts that turn GPT-Image2 into a predictable image-making tool
- block/buzzAn open-source workspace where humans and AI agents chat, code, and review in the same rooms
- NousResearch/hermes-agentNous Research's Hermes is an AI agent that gets smarter the more you use it
- virgiliojr94/book-to-skillA tool that turns technical book PDFs into on-demand reference skills for AI coding agents
- VoltAgent/awesome-agent-skillsA single hub collecting over 1000 'how-to' manuals that make AI coding assistants act like experts
- anthropics/claude-plugins-communityA shared shelf where anyone's Claude add-ons get listed for install