Flipbook is an infinite visual browser where every page is a real-time AI-generated image; users navigate by clicking anything in the image to go deeper.
Key Takeaways
All text and UI elements are rendered as pixels by the image model; there is no HTML, CSS, or overlay layer.
Clicking any region of a generated image triggers a new, deeper-dive image rather than following a conventional hyperlink.
Content is grounded via agentic web search plus the image model’s world knowledge, at roughly ChatGPT/Gemini/Claude accuracy levels.
An experimental live video stream feature animates pages and builds transitions using a separate custom video generation model, kept behind a toggle due to high resource cost.
Roadmap targets real data integration, in-image interactivity, action-taking, and eventually replacing separate apps with Flipbook-native workflows.
Hacker News Comment Review
Commenters validated factual accuracy in narrow technical domains, with one user successfully generating correct suspension torque specs and component diagrams for their specific car.
A persistent open question is how session context and page consistency work when every render is generated fresh, including whether previously visited pages can be revisited reliably.
Inference cost is a visible concern: the public demo offers no transparency on GPU spend or subsidy, which commenters from constrained backgrounds found difficult to reason about.
Notable Comments
@nirav72: uploaded an external hydroponics diagram as input; Flipbook generated detailed sub-diagrams for plumbing, nutrient delivery, and electrical wiring from it.
