How it works

Point, segment, tap.

Amun does not type glyphs for you. You photograph the stone. Models find the pieces. You choose what to read.

Point and capture

Aim at a museum piece, a site wall, or pick a photo from your library. The same flow works on the Djed-Hor demo statue in onboarding.

AI segmentation

The neural pipeline detects individual glyphs, cartouches, and artifact boundaries, including worn or partly damaged surfaces when it can.

Read and translate

Tap a detected region. Amun returns transliteration, translation, and historical context matched against curated databases where those exist.

What happens on the server

The photo is uploaded so a GPU worker can run detection and retrieval. A Firebase job record tracks status. That is why a live connection is required for new scans. Results can then be saved to your archive on the device and in your account.

Try it without a museum

The first-run tutorial can process the Statue of Djed-Hor (Cairo Museum, c. 310 BC), a healing statue covered in magical texts. That is the same object shown on the homepage.