Aria

What Are AI Calories? FAQ

People hear AI Calories and assume the app invents a calorie number from a photo. That is not how AuraBase works. Below are the short answers we give when someone asks what AI Calories actually are — and how the AI Meal Scanner produces them.

What are AI Calories?

In AuraBase, AI Calories are the calorie total for a logged meal after the AI Meal Scanner has identified the foods and portions, then looked those items up in official nutrition databases. The energy value comes from published reference data (per 100 g or per serving), scaled to the estimated grams on your plate — not from a language model making up a nutrition label.

How does the AI Meal Scanner extract AI Calories?

Four steps, on purpose:

Snap — You photograph the meal.
Identify — AI names the foods and estimates portion size in grams.
Look up — Each item is matched to a row in USDA FoodData Central and/or the Canadian Nutrient File (CNF).
Scale — AI Calories (and macros/micros) are calculated from that reference row × estimated grams.

Vision finds what is on the plate. The database supplies what is in each gram. That split is the whole point of calling them AI Calories in our product — AI helped get the meal into the log; the calorie math is still table-backed.

Are AI Calories the same as an AI “guessing” calories?

No. A vision-only app can spit out a calorie total straight from the model’s training priors. That looks fast, but the number is a guess dressed like a lab result. AuraBase does not ask the model to invent calories. The AI Meal Scanner stops at identification and portion; AI Calories come from the matched USDA/CNF entry. If you change the portion slider after a scan, the calorie total updates from the same fixed composition row — it does not re-roll a new fiction.

Why USDA and CNF?

Those are government reference databases used by researchers and dietitians for food composition. We use both because people eat globally — a dish may match better in one table than the other. AI Calories only mean something if you can say which reference row they came from.

Can I trust AI Calories completely?

Trust the composition more than the camera. Portions estimated from a single photo can still be off — sauces, hidden oils, and mixed plates are hard for any scanner. What you should expect from AuraBase: when the food match is right, the calorie density per gram is from a real database row. Review the ingredient list, nudge the grams if you know better, then log. For the research behind this approach, see What Research Says About AI Meal Photo Accuracy and Inside AuraBase's Hybrid AI Meal Scanner V2.

Quick takeaway

AI Calories in AuraBase = AI Meal Scanner finds the food and portion + USDA/CNF does the calorie math. Same idea as scanning a barcode for packaged food, except the “label” is a government reference entry matched to what was on your plate.