Snapkin launches on iPhone and Android: calorie tracking that takes one photo, measured against 500 weighed meals
The Belgian app replaces the scale, the barcode scanner and the food database with a single photograph, and publishes how accurate it is.
Snapkin, a new calorie and protein tracker, is now available on the App Store and Google Play. It is built for the people who have tried calorie tracking before and stopped because it asked too much of them: weighing rice, searching long food lists, scanning barcodes, and typing in every ingredient. With Snapkin, the user photographs the plate. The app does the rest.
Built for the people who gave up on tracking
Most people who start logging their food stop within weeks. In one published study of a free photo food-logging app, only 2.58% of the 189,770 people who opened it went on to use it actively (Helander et al., 2014). Snapkin's founder believes the reason is effort, not willpower.
Snapkin removes that effort:
- One photo logs a meal. The app names each item on the plate, estimates each portion from the plate and cutlery beside it, and counts the cooking oil and sauces that hand-logged meals usually miss.
- It asks only when it matters. When one detail would change the number, such as the oil a dish was fried in, Snapkin asks one question and shows how much the answer matters. "I do not know" is always an option.
- It learns the user's kitchen. Every answer is saved as a plain sentence the user can read and delete. After a few weeks, everyday meals are logged without questions.
- It shows its reasoning. Every ingredient shows what the app assumed, and every meal says how certain the estimate is.
- It fits into the day. Home Screen and Lock Screen widgets, weight read automatically from Apple Health and Health Connect, restaurant menu estimates, and suggestions for what to eat next.
- It speaks the user's language. Ten languages, including the questions the app writes about each meal: English, Chinese, Hindi, Spanish, French, Bengali, Portuguese, Indonesian, German and Dutch.
Accuracy, measured and published
Snapkin was tested against 500 meals from three published research datasets, where every meal had been weighed on a scale and photographed. The test used a single photo per meal, before any follow-up question was answered.
- On phone photos, Snapkin reached 83% median accuracy per meal. Median accuracy is 100% minus the median error: on a typical meal, Snapkin's calorie estimate was within 16.8% of the weighed value. Across all three datasets, the figure was 80%.
- Over a simulated week of 21 meals, median accuracy rose to 90% on phone photos, and 89% across all three datasets, because small errors on single meals cancel out in the total.
- Large errors are rare. 6 of 225 phone-photo meals were off by more than half.
For context, the researchers behind one of the datasets reported mean errors of 23% to 34% for four general-purpose AI models on its photos. Snapkin's mean error on the phone photos was 19.1%. The two tests used different instructions, photos and reference values, so this is not a direct comparison.
Before choosing the AI model that runs the app, the team compared thirteen models across 3,195 test runs, scored against meals that had already been corrected by hand. The same 500-meal test is run again before any change to the model or its instructions, and the published figures are updated whichever way they move. The full method, the limits of the test, and every figure are published at getsnapkin.app/en/research/own-accuracy.
Research behind every decision
Beyond accuracy, Snapkin publishes the reasoning behind each product decision on getsnapkin.app/en/research: how calorie targets are calculated, why protein is treated as a minimum, how fast the app lets a user lose weight, and why exercise does not add calories to the daily budget. The pages cite 88 sources and include the evidence against the team's own choices, including a page on the relationship between calorie tracking and eating disorders and what the app deliberately refuses to do.
Founded by an infrastructure and reliability engineer
Snapkin is built by Mattias Geniar, a Belgian engineer with nearly two decades of experience running software that people depend on. He spent twelve years at Nucleus, a Belgian 24/7 managed hosting provider, and co-founded Oh Dear, a website monitoring service used to watch uptime, security certificates, DNS and broken links for sites around the world. He has written about Linux, security, performance and the web since 2008, and has spoken at conferences including FOSDEM and Laracon EU.
That background shapes how Snapkin handles personal data. User data is stored in the European Union. The app shows no advertising, has no tracking pixels, and does not sell food diaries or use them to train anyone's AI models. Photographs are stored privately and are deleted with the meal.
"Every calorie tracker I tried worked, technically. Then a busy week came, I could not face weighing rice again, and I stopped. I did not need more discipline. I needed an app that asked less of me. Snapkin takes one photo, asks a question only when the answer changes the number, and learns the rest. And because an estimate is only useful if you can trust it, we measured it against weighed food and published every number."
Mattias Geniar, founder of Snapkin
Availability and pricing
Snapkin is available now for iPhone (iOS 17 and later) and Android. One subscription includes every feature, with no photo limits: €9.99 per month with a three-day free trial, or €99.99 per year. Prices are billed by the App Store and Google Play in local currency.
Snapkin is a general wellness tool and is not a medical device.
About Snapkin
Snapkin is a photo-based calorie and protein tracker for iPhone and Android, developed by Robot Studios IT Consultancy BV in Berlaar, Belgium. More information, research and press materials are available at getsnapkin.app.
Media contact
Mattias Geniar
[email protected]
getsnapkin.app
###