Guide
How to count calories from a photo, and how accurate it really is
A photo calorie counter works in three steps: the AI identifies the foods, estimates the grams of each, then converts grams to calories and macros. Accuracy depends mostly on step two (portion size) and on whether step three uses a real nutrition database or the AI's guess. Kalova uses USDA FoodData Central for that step when a food matches.
How does counting calories from a photo work?
- Identify the foods. A vision model looks at the photo and names what it sees: rice, grilled chicken, salad.
- Estimate grams. The model estimates how much of each food is on the plate. This is the hardest step, because a photo is flat.
- Look up nutrition. Grams are converted to calories, protein, carbs and fat. Either the app looks each food up in a nutrition database, or the AI model outputs the numbers itself. A database number can be traced to a source; a model's number cannot.
How do you count calories from a photo in Kalova?
- Open Kalova and take a photo of your meal, ideally from above.
- The AI lists the foods it found and estimates the grams of each.
- Kalova takes calories and macros from USDA FoodData Central for each food that matches. Anything without a match is marked as an estimate.
- Open the Nutrition Receipt to see every item with its grams and source. Edit the grams if they look wrong, then save.
The same photo gives the same answer every time. If you would rather not take a photo, you can describe the meal in words, dictate it (speech recognition runs on the iPhone), scan a barcode, search for a food, or scan a nutrition label. The full method is on the methodology page.
How accurate are photo calorie apps?
Honestly: good enough to track trends, not good enough to replace a kitchen scale. The error comes from three places, in roughly this order:
- Portion size. Estimating grams from a single 2-D photo is the main error source for every photo app, Kalova included. Depth sensors on some iPhones help, but no app can see what is under the top layer.
- Hidden ingredients. Oil, butter, sugar and sauces are often invisible.
- The nutrition data. If the numbers come from the AI rather than a database, they can drift between scans. This error is the one you can avoid by choosing the app.
Accuracy is also different by meal type. Here is how the main error sources apply:
| Meal type | Portion estimate | Hidden ingredients | What to check |
|---|---|---|---|
| Single item (apple, slice of bread) | Easier | Few | Grams |
| Separate items on a plate (meat, rice, vegetables) | Moderate | Cooking oil | Grams of each item, added oil |
| Mixed dish (stew, curry, casserole) | Harder | Oil, cream, sugar | Overall portion, main ingredients |
| Restaurant meal | Harder | Often several | Portion, oil and sauces |
| Packaged food | Use the barcode or label instead | Listed on the label | Serving size |
How can you get better calorie estimates from a photo?
- Shoot from above. A top-down photo shows the whole plate and the edges of each item.
- Include something for scale. A fork, a hand or a standard plate helps judge size.
- Separate the components when you can, rather than photographing a mixed pile.
- Check the grams before saving. This is where most of the error lives.
- Log oils and sauces by hand. The camera usually cannot see them.
How do you test your current photo calorie app?
Run the five-scan test. Plate one meal, leave it untouched, and scan it five times from the same angle. If you get one number five times, the app is consistent. If you get five different numbers, it is re-guessing each time. Then open a result and look for where each number came from. We will publish our own results for several apps when we have run the test.
Limitations
- No photo app, including Kalova, can reliably measure portion size from one 2-D photo. Treat every result as an estimate and correct the grams.
- Kalova's USDA lookup covers foods that match a USDA entry. Items without a match are marked as estimates.
- Kalova is iPhone only (iOS 17 or later). There is no Android version.
- Kalova is not a medical device. If you manage a medical condition with your diet, work with a clinician.
Frequently asked questions
Can AI see oil and butter in a photo?
Usually not. Cooking oil and butter soak into food or sit under it, so a photo often cannot show them. A tablespoon of oil adds a meaningful number of calories, so add cooking fat by hand when you know it was used.
Does photo calorie counting work for homemade dishes?
Yes, especially when the components are separate on the plate. Mixed dishes such as stews and casseroles are harder, because the app cannot see what is underneath. Check the grams of each item and correct them before saving.
Do I need an account to count calories from a photo in Kalova?
No. Kalova works without an account, email or sign-up, and the food log stays on your iPhone.
Is my meal photo stored?
In Kalova, the photo is sent through Kalova's server to a third-party AI provider for analysis and is not stored server-side. Other apps have their own policies; check the App Privacy section of their App Store listing.
Why do some apps give a different number when I scan the same meal twice?
If the AI model produces the calorie number directly, small differences in how it reads the photo can change the result. Apps that look the food up in a database and keep the same identification give the same answer for the same photo. The five-scan test shows which kind you have.