Reference data

650+ foods, curated by hand.

Meal Scanner's text mode, photo mode, and combo builder all resolve to the same local food dataset — a hand-curated table, not a live lookup — so a meal typed twice, a year apart, still scores identically.

20 categories, two cuisines' worth of default coverage

The dataset leans deliberately Indian-first, with dedicated categories most Western nutrition apps skip entirely, alongside standard Western and global staples:

Indian breadsIndian riceDals Indian veg curriesIndian non-veg South IndianIndian snacksIndian sweets Indian breakfastIndian beverages

Western proteinsWestern carbsProduce DairyNuts, seeds & oilsWestern fast food Global cuisineWestern snacksCondiments Western beverages

What a food record holds

Same shape whether it's dal makhani or a protein bar: a name, calories, and protein / carbs / fat in grams — the same four numbers the Dashboard's macro bars are built from, so nothing has to be converted or re-scaled between screens.

Fuzzy matching, on purpose

A typed or photo-identified dish name doesn't need to match a record exactly — the same fuzzy matcher scores "chicken tikka" and "chicken tikka masala" close enough to resolve correctly, which is what lets photo mode and text mode share one matching path instead of needing separate logic.

Packaged food, live

Anything with a barcode steps outside the curated dataset entirely: Barcode mode queries Open Food Facts directly, so a packaged product's real label data — including sugar and trans fat, plus its Nutri-Score and Green-Score grades — comes back live rather than from a static table.