How AI Menu APP reads a menu

The process begins with the source that best preserves the restaurant's information. A digital PDF is usually preferable to a screenshot because it can retain sharper text and complete pages. For a printed menu, a straight photo with even lighting reduces missing lines and incorrect characters.

A public link may contain text, a linked PDF, or images. Some sites load content dynamically or block automated access, so a file upload can be more reliable when a link cannot be read.

Text extraction identifies visible words while the parser separates restaurant information, headings, dish names, descriptions, and prices. Layout matters because a number beside a dish is usually a price, while a large label above several dishes is more likely a category.

Menus often mix languages, currencies, serving sizes, and decorative text. Language profiles and general menu patterns help, but unusual typography or missing context can still produce the wrong grouping.

Dish names and descriptions are translated after the structure is recovered. Familiar culinary terms may be preserved when a literal translation would be misleading, while descriptions provide context for ingredients and preparation.

Regional names, handwritten changes, and culturally specific dishes may require clarification from restaurant staff. Keep the original wording available so staff can identify the exact item.