This page
A static file served from GitHub Pages. It has no backend, no database and no upload endpoint. It cannot receive an image; there is nowhere for one to go.
Runs on your machine · nothing is uploaded
Choose a folder of album pages, watch each one being separated, and download the photographs as a zip. The page you are reading is a static file with no server behind it. The work is done by a copy of RevelAI running on your own computer, which you start and stop yourself.
This page has no server behind it and never receives an image. It drives a copy of RevelAI running on your own machine, so your photographs are read by a program on your computer and go nowhere else. Two commands, once:
Install
pip install revelaiStart it
revelai serve --openserve listens on 127.0.0.1:8765, which is this computer and nothing else — no other machine on your network can reach it. Leave that terminal open while you work, and stop it with Ctrl+C when you are done. Everything it wrote is deleted when it stops.
Running it on another port? Add ?port= to this page’s address.
What is actually happening
A static file served from GitHub Pages. It has no backend, no database and no upload endpoint. It cannot receive an image; there is nowhere for one to go.
A small server you start yourself, bound to 127.0.0.1. That address is your own machine and is not reachable from your network, let alone the internet. It refuses requests from any page but this one.
The same Python code the command line runs, doing the same work with the same output. The browser is a remote control for it, not a replacement.
http://127.0.0.1:8765, which is a program on your own computer. They are written to a temporary folder that is deleted when you stop the server.$ revelai serve --openRevelAI 0.1.0 - split album pages, then restore the photographs Listening on http://127.0.0.1:8765Open: https://maikotrindade.com/revelAI/run/Settings: backend: local (local only) | faces and colourisation: refused Your photographs are processed here, on this machine, and are notuploaded anywhere. Press Ctrl+C to stop.Before you start
holiday.jpg that is really a spreadsheet is caught here..DS_Store, Thumbs.db and their relatives are skipped rather than refused. Your desktop put them there; that is not you doing anything wrong.A zip holding images only, numbered continuously — photo_00000001.png onwards. No logs, no JSON, no thumbnails, no folders. Every file in it is a lossless PNG carrying the source page’s colour profile and a record of what was done to it.
Pages RevelAI is not confident about are listed on the results screen by name and reason. They are still written — a flagged crop is usually the right crop — but you are told, which is the whole point.
Working through a big album? Read how to photograph album pages first. It makes more difference than any setting here.
Or don’t use a browser
Split only
revelai split ./album-pages -o ./photosSplit, then restore
revelai run ./album-pages --out-split ./photos --out-enhanced ./restoredRestore only
revelai enhance ./photos -o ./restored --compare-dir ./compareServe this page’s engine
revelai serve --openThe full flag reference, the detection write-up and the honest limitations are in the engine README. The limitations are worth reading before you restore anything you care about: face restoration reconstructs faces rather than revealing them, and colourisation invents the colour.