TranslatAR
Point and translate in place. Menus, signage, small print, painted walls.
What it does
Point the camera at text in another language. The instrument reads it, translates it, and writes the translation back over the same place the original sat. Tilt the phone — the translation tilts with the surface, because it's anchored to the surface, not the screen.
— the translation lands here —
How it works
Apple Vision OCR runs on each frame. The detected text blocks are dispatched to a translation engine — local for the languages Apple supports natively, server-side for the rest. The result is composited back via the AIM (anchored interface marker) overlay math: the text renders on the same plane as the source, follows the same homography, occludes correctly.
What it's built on
Apple Vision OCR. Apple Translation framework for the supported pairs. A server-side translator for the rest. The AIM overlay math is custom — it's why the translation sits on the surface and not floating in front of the phone.