OCR for Menus and Documents, in Arabic and English.
Our OCR turns photos of menus and printed documents into structured data. It is in the demo stage: the video below is a real run on a fictional menu, and we can run the demo on one of your menus or documents.
A real run of the app on a fictional menu we printed for the demo. Muted loop.
Four steps, from photo to file.
Photograph And Upload
The menu is photographed straight-on with a phone and uploaded: one page, a board or a flyer.
Read Up Close
The page is read in overlapping zoomed parts, together with a scan of its sections and price columns.
Double-Check
Disputed or uncertain items are read again on a close-up, and every section count is cross-checked.
Review And Export
A person checks the flagged items, edits anything in the table, then downloads the file.
Built for the menus restaurants actually print.
Built For Arabic Menus
Right-to-left columns, Eastern-Arabic digits and dishes printed in both languages are read as one dish with both names.
It Shows Its Doubts
An item it is not sure about is flagged with the part of the photo it was read from, so a person checks only the rows that need it.
Files Your Team Can Use
Excel, CSV or JSON. A direct connection to your POS or ordering system is a custom build we scope with you.
What the result looks like.
What people usually ask first.
Not yet. Our OCR is in the demo stage, which is why this page makes no accuracy claims. Try the demo on your own menu or document first.
Printed menus, boards and flyers photographed with a phone, in Arabic, English or both. Very small or blurred photos are marked as low quality before you rely on them.
Check the flagged items. Each one comes with its part of the photo, so it is quick.
Menus first. We use the same reading approach for other printed documents; ask us to run the demo on one of yours.
Not out of the box. Writing items straight into a POS or ordering system is a custom connection. Tell us which system you use and we check what its API supports.
There are no public prices for OCR yet. Request a demo on your own menu or document, and we talk about your volume after you have seen the result.