Atlaso started with a reel. My girlfriend sent me one on Instagram while we were planning a trip, basically saying we should order one of those travel photobooks to remember it by. I looked at the prices, looked at how clunky most of these services were, and thought, “I could just build this.” So instead of buying the book, I went and built the company.
The pitch was simple: dump all your travel photos in, let AI do the boring part, and get a print-ready hardcover book out the other end. No curating, no fiddling with layouts, no design skills required.
Under the hood, it turned into a surprisingly deep project. You upload a whole trip’s worth of photos and the backend runs each one through GPT-4 vision to score it on things like sharpness, aesthetic quality, scene type, faces, lighting, and dominant colours. From there a selection algorithm whittles hundreds of photos down to the best ~30. It deduplicates burst shots, balances the mix between landscapes, people, food and city scenes, and orders everything chronologically so the book actually reads like the trip happened.
Then it groups those photos into spreads by looking at time gaps, location changes and visual similarity, picks a layout template for each page, and renders the whole thing into a print-ready PDF. The frontend handles the rest of the journey: drag-and-drop upload with EXIF and location detection, client-side HEIC conversion for iPhone photos, a cover designer with templates and colour pairings, a live preview where you can drag photos around to reframe them, and a full order and checkout flow.
I built both sides end to end. The backend is Kotlin and Spring Boot with Postgres, the frontend is Next.js and React, with Google sign-in, S3 uploads, and OpenStreetMap for reverse geocoding the trip location from photo GPS data.
It worked. It genuinely worked, and the output looked good. But I never took it live. The more I sat with it, the clearer it got that the hard part was never the software. It was the operations: managing a print partner, quality-checking physical books, handling shipping and returns, customer support when a corner arrives bent. That is a logistics business wearing a tech costume, and I wasn’t excited to run it. So Atlaso stayed a really polished thing I built for myself and one very specific trip.
No regrets though. It was one of the most complete products I’ve built solo, and it scratched a real itch.
The preview editor, where the whole book comes together and you can drag photos around to reframe them:
One of the cover templates, with a trip title and colour pairing:
And the generation screen, where the AI does its thing:
