Count calories from a food photo with AI

How accurate is it?

Modern AI assistants recognise most dishes well and estimate portions from the plate, cutlery and packaging. The usual error is in what they can't see: oil in the pan, sauce under the food, butter in mashed potatoes. Expect the total to be off by 10–25% on a mixed plate, and much less on packaged food with a visible label.

How to get better estimates

• Shoot from above with the whole plate and a fork or a hand in the frame for scale.
• Mention what the photo can't show: "fried in a tablespoon of oil", "200 g of rice".
• Photograph the nutrition label of packaged food: the numbers per 100 g are exact.
• For a dish you cook often, give the recipe once and the cooked weight; then log it in grams.

Turn photos into a food log

In LiftBrain you take the photo in the app, even at the table without opening a chat. The photo waits for your AI assistant; next time you talk to it, it reads the photos, logs each meal with calories, protein, fat and carbs, and shows the day against your targets. Your training plan and your food live in one place, so the coach can tell you to hold the weight on a day you ate too little.