Context
Lovetopus is a series of picture books for children aged 3 to 6. Each story is built around one emotion and set in an undersea world called EmOcean. The printed book is the product; a web zone for parents is planned to accompany it.
A series needs a production line, not a one-off layout job: consistent characters across books, pages that meet print specifications every time, and a way to link each printed copy to material online without reprinting when that material changes.
What we built
- Book Tool. A print-production engine with an on-page editor. Text is edited directly on the rendered page; each spread moves from draft to approved; a print gate refuses to produce files until the print specification is met. Output is print-ready for a square 8.5 × 8.5 in format.
- Illustration pipeline. Illustrations generated with image models, checked by independent automated reviewers against the character canon, and cut out from their backgrounds. Final pages are composed and approved by hand.
- QR registry. Each printed QR code has a fixed token and an editable destination, so the target can change while the printed code stays the same. Anonymous scans are counted only in aggregate.
- Back end. Cloudflare Workers with a D1 database; administration sits behind access control.
- Website. lovetopus.com on Cloudflare Pages, with a headless WordPress blog and transactional e-mail.
How we built it
The Book Tool runs as a local application: a Python engine renders pages to print-ready PDF, and a React and TypeScript editor works on top of the same rendered output, so what is edited is what is printed. The book's content lives in a structured model rather than in layout files, which keeps every book in the series consistent.
Specifications were reviewed by an independent critic before build, and the engine ships with its own automated tests. Every generated illustration passes an independent check before it can be used: the agent that produced an image never approves it.
Where AI is used
- In the product
- Illustration generation with image models, followed by independent automated quality checks and background removal. Text and final page composition are human.
- In how it was built
- An AI-agent engineering team with independent read-only reviewers, under human approval.
Where it stands
The platform is in development. The first book is being produced, the website and QR registry are running, and the parents' zone is planned. We build and operate the platform under licence.