Food scanner for calories: every kind of food, and how to scan it

Three kinds of food, three methods

The instinct to use one method for everything is what causes trouble. Photographing a cereal box when the barcode is right there throws away a precise answer in favour of an estimate.

Packaged food: scan the barcode

If it has a barcode, that is always the best route. You get the manufacturer's own declared panel rather than anyone's estimate, specific to that exact product and that exact formulation.

The one habit that matters: check the portion. Databases store per 100 g, but you are eating a portion. Our calorie barcode scanner defaults to the manufacturer's serving, shows what it weighs, and recalculates as you adjust it.

Watch for multipacks. Scanning the outer box of a six-pack of yoghurts may return figures for the whole box rather than one pot. The stated serving size tells you which you are looking at.

Cooked meals: photograph the plate

Most of what people actually eat has no barcode. A curry, a canteen tray, a plate at a friend's house — none of it carries a label, and reconstructing it from ingredients is slow enough that most people give up.

The AI calorie scanner estimates each item's portion weight from the photograph and derives the macros from that. Shoot at an angle rather than straight down, get a fork or a standard plate in frame for scale, and add a note for anything the camera cannot see — cooking oil above all.

Fresh produce: look it up by name

An apple has no barcode and does not need a photograph. Search the name and take the standard reference value, which for unbranded whole foods comes from USDA FoodData Central and is about as settled as nutrition data gets.

The variable here is weight, not composition. A medium apple is 180 g and a large one is 250 g — a 40% difference on the same food. If you weigh fresh food for a week, you will calibrate your eye permanently and never need to weigh it again.

Two traps are worth naming. First, raw versus cooked: 100 g of dry pasta becomes around 250 g cooked, and the calorie figures differ accordingly. Match the state of the food to the state in the entry. Second, bone and skin: chicken thigh figures vary hugely depending on whether the entry includes them.

The awkward cases

Restaurant meals

Chains often publish nutrition data, which beats any estimate — check their site first. For independent restaurants, photograph the plate and expect the estimate to run low, because restaurant cooking uses considerably more fat than home cooking. Adding 15% to a restaurant estimate is a reasonable working correction.

Drinks

Scan the bottle or can. This is exact, and it matters more than people expect: drinks are easy to consume in quantity and easy to forget entirely. Alcohol is the common blind spot at roughly 7 calories per gram, second only to fat.

Cooking oil

The most under-recorded item in home cooking. A tablespoon is around 120 calories, and what goes in the pan does not all end up in the food — but a good deal of it does. Measure it rather than pouring, at least until you can judge it.

Deli-counter and weighed items

Barcodes starting 02 or 20–29 are shop-internal, encode a price, and mean different things in different branches. There is nothing to look up. Search the food by name instead.

How much precision is worth having

A useful principle: precision should match the size of the thing being measured. Getting a 600-calorie dinner roughly right matters enormously. Deciding whether a stick of celery is 6 or 9 calories does not, and the time spent on it is the main reason people abandon tracking altogether.

Scan what is easy to scan, estimate the rest, and let the week's pattern do the work rather than any single entry. That is a tracking habit that survives contact with real life, which is the only kind that helps.

Open the scanner — your first scan is free.