Does it work on mixed dishes or restaurant food?

Yes, with more uncertainty. Plates where foods sit side by side scan best. For stews, curries, casseroles, burritos and restaurant meals, the scanner names the likely components and estimates portions, but sauces and cooking fat are guesses, so check the portions and add the oil or butter you expect the kitchen used.

No. The photo is resized in your browser, sent over an encrypted connection, analysed in memory by the AI model and then discarded. FitStack keeps no copy and returns no image link. To enforce the free daily limit we keep only a count of scans per day, keyed to a one-way hash of your connection and a random browser id.

Is the AI meal scanner free?

Yes. You get 1 free scan a day without signing up and 3 a day once you add your email. Inside the FitStack app, meal scanning is part of everyday food logging.

Does it work offline?

No. The analysis runs on FitStack servers, so the web scanner needs an internet connection. If a scan fails or times out on our side, it is not counted against your daily limit.

Photo, barcode or manual logging: which should I use?

Scan the barcode for anything packaged, because the label is more accurate than any photo estimate. Use a photo for home-cooked or restaurant plates with no label. Log manually, ideally with a food scale, when accuracy matters most, such as the first weeks of a diet or a meal you eat every day.

Free tool

AI Calorie Counter: Log Meals From a Photo

Take a photo of your plate and the scanner lists each food it sees with an estimated portion, calories, protein, carbs and fat. Every line is editable, because a photo is a starting estimate, not a weigh-in.

By the FitStack team · Last updated October 18, 2026

Photograph your plate

Shoot from above in good light with the whole plate in frame. On a computer you can drag a photo here.

Take a photoUpload a photo

Free: 1 scan a day, 3 with your email. Your photo is resized in your browser, analysed and then discarded; it is never stored.

How the photo calorie counter works

Step 1

Snap the plate

Shoot from above in good light. Your browser shrinks the photo before upload so it sends in a second or two on mobile data.
Step 2

AI itemizes the meal

A vision model names each food and estimates its weight. Where a food matches a label FitStack has reviewed, those label values are used instead of a generic estimate.
Step 3

Check and correct

Adjust any portion in grams, add the oil or butter the camera missed, and the totals update instantly.

How accurate are AI calorie counters?

Photo apps are convenient, but they miss calories in predictable ways. A 2026 study run by NIH researchers tested four popular photo-based apps against measured meals and found they underestimated calories by roughly 250 to 345 kcal per meal and fat by about 30 g, close to a third of the meal, with high-fat keto meals the hardest to judge.^[1]^[2] The results were presented as a conference abstract and have not yet been peer reviewed, but the direction matches what dietitians see: the missing calories are mostly fat.

That is why this scanner shows its working instead of a single number. You see every food and portion it assumed, so a wrong guess is obvious and takes seconds to fix. We don't quote an accuracy percentage: we haven't run a reproducible test against weighed meals yet, and a number without a method would not mean much.

Where photo AI gets it wrong

Cooking oil, butter and dressing. One tablespoon of olive oil is 119 kcal and is invisible once it's in the pan.
Portion depth. A single photo shows area, not height, so a deep bowl of rice and a shallow one can look alike.
Mixed and layered dishes. Lasagne, burritos, curries and casseroles hide what's inside.
Look-alikes. Whole and skim milk, regular and diet soda, rice and cauliflower rice photograph the same.

Tips for a better estimate

Photograph from directly above, with the whole plate in frame and nothing stacked on top.
Include a fork or your hand for scale, and split foods apart on the plate when you can.
Add the cooking fat and sauces you know were used with the one-tap add-ons.
Correct the portion of the biggest calorie item first; it moves the total the most.

Photo vs barcode vs manual logging

When to use each way of logging food
Method	Best for	Speed	Main source of error
Barcode scan	Anything packaged	Seconds	Eating a different amount than the label serving
Meal photo	Home-cooked and restaurant plates	Seconds, plus a quick check	Hidden fats and portion depth
Manual with a scale	Meals you repeat, early weeks of a diet	A minute or two	Picking the wrong database entry

For packaged foods, look the product up in the food nutrition facts directory: it uses the label itself.

Your photo is not stored

The photo is resized and re-encoded in your browser before upload, which drops its location and camera metadata.
It is sent over HTTPS, analysed in memory and discarded. FitStack keeps no copy and no image link.
To enforce the free limit we store only a daily scan count against a one-way hash of your connection and a random browser id.
Your email is only used if you give it, for what you asked for. See the privacy policy.

Frequently asked questions

How accurate are AI calorie counters?: Less accurate than their marketing suggests. In a 2026 test of four popular photo apps reported by Healio and HealthDay, the apps underestimated meal calories by roughly 250 to 345 kcal and fat by about 30 g, around a third of the meal, with keto-style meals the hardest. That work was a conference abstract, not yet peer reviewed. A single photo cannot see oil, butter or what is under the top layer, so FitStack shows every item and portion for you to correct and lets you add cooking fats in one tap. We do not publish an accuracy percentage because we have not run a reproducible test yet.
Does it work on mixed dishes or restaurant food?: Yes, with more uncertainty. Plates where foods sit side by side scan best. For stews, curries, casseroles, burritos and restaurant meals, the scanner names the likely components and estimates portions, but sauces and cooking fat are guesses, so check the portions and add the oil or butter you expect the kitchen used.
Is my photo stored?: No. The photo is resized in your browser, sent over an encrypted connection, analysed in memory by the AI model and then discarded. FitStack keeps no copy and returns no image link. To enforce the free daily limit we keep only a count of scans per day, keyed to a one-way hash of your connection and a random browser id.
Is the AI meal scanner free?: Yes. You get 1 free scan a day without signing up and 3 a day once you add your email. Inside the FitStack app, meal scanning is part of everyday food logging.
Does it work offline?: No. The analysis runs on FitStack servers, so the web scanner needs an internet connection. If a scan fails or times out on our side, it is not counted against your daily limit.
Photo, barcode or manual logging: which should I use?: Scan the barcode for anything packaged, because the label is more accurate than any photo estimate. Use a photo for home-cooked or restaurant plates with no label. Log manually, ideally with a food scale, when accuracy matters most, such as the first weeks of a diet or a meal you eat every day.

Scan every meal in the FitStack app

In the app, a scanned meal goes straight into your food log, next to barcode scans and your workouts, and your calorie target recalibrates from what you actually eat and weigh.

Launching on iOS and Android.

Sources

Add-on and protein values (olive oil, butter, mayonnaise, chicken, yogurt, eggs, tofu) are from USDA FoodData Central.