AI calorie counter for iPhone

An AI calorie counter estimates the calories and macronutrients in a meal from a photograph, instead of asking you to find each ingredient in a food database and set its portion by hand. You point the camera at your plate, the model identifies what is on it, and the calories, protein, carbohydrates and fat land in your diary.

Frulo is an AI calorie counter built for iPhone. This page explains how the technology works, where it is genuinely good, where it is not, and how to decide whether it fits how you eat.

Why photo-based calorie counting exists

Calorie tracking works. The reason most people stop is not that the arithmetic fails them — it is that logging a meal takes ninety seconds of searching, scrolling and correcting portion sizes, three times a day, forever. A week of that is fine. A month is a chore. Six months is unusual.

Photo-based logging attacks the friction rather than the maths. One photograph replaces the search, the disambiguation between fourteen near-identical database entries, and the portion dropdown. The estimate that comes back is less precise than a weighed and hand-entered meal — but a rough number you actually record beats an exact number you never get around to entering.

How the estimate is produced

When you photograph a meal in Frulo, the picture is downscaled and sent for analysis. A vision model identifies the distinct foods on the plate, estimates how much of each is present relative to the plate and other reference objects in frame, and maps those to nutrition values. What comes back is a per-food breakdown you can review, edit, and then save.

The photo itself is never kept. It is analyzed and discarded — not saved on your device by Frulo, not stored on our servers, and not sent to analytics. Only the resulting numbers stay, in your own diary.

How accurate is an AI calorie counter?

Honestly: it is an estimate, and it is worth understanding its limits before you rely on it.

A photograph cannot show you everything. It cannot see the tablespoon of oil a vegetable was roasted in, it cannot see what is underneath the top layer of a mixed bowl, and it cannot tell full-fat yoghurt from the low-fat version in the same white dish. Those are not shortcomings of one app; they are what is missing from the image. Every photo-based calorie counter shares them.

What makes this workable is that body composition responds to your average intake across weeks, not to the precise figure on any individual meal. Errors in both directions across dozens of meals largely cancel, and the trend line — which is the thing you are actually steering by — stays informative. That is why Frulo puts weight trend and calorie averages on the statistics screen rather than celebrating a single perfect day.

When you do want precision, you have exact routes: scan the barcode of a packaged food and the real nutrition panel is used verbatim, or type the food and portion yourself.

When to use a photo, and when not to

A photo is the fastest option for:

Reach for something else when:

What Frulo adds around the scanner

A scanner on its own is a party trick. The parts that make it a tracker you keep using:

Privacy, in one paragraph

Meal photos are analyzed and discarded, never stored. Frulo contains no advertising SDK, never asks for the tracking permission, does not sell your data and does not share it for advertising. The AI provider that processes scans and coach messages does so as our processor and does not use them to train its models. Anything Frulo holds you can delete yourself, up to and including the whole account, from inside the app. The full detail is in the privacy policy.

Try it

Frulo is free to download and starts with a free trial. It runs on iPhone; there is no Android or web version yet.

More detail: the full feature list, or the frequently asked questions.