CountNutri

Snap a photo of your meal — AI counts the calories

Health & Fitness

Visit WebsiteSee on Product HuntFacebook ⧉

Hi Product Hunt 👋 I'm Sampath, from Sri Lanka. I built CountNutri because every calorie tracker I tried failed at my own dinner. Our food arrives as mixed curries on one plate — rice, dhal, a coconut sambol, a jackfruit curry. Logging "white rice, 1 cup" while everything else goes uncounted isn't tracking, it's guessing. And building a custom recipe entry for every meal is exactly how people quit in week two. So CountNutri works the other way round: photograph the plate, and the AI identifies each dish, estimates the portions, and returns calories plus protein, carbs and fat, cross-checked against USDA data. Any item can be corrected in a tap. The honest part: recognition is the easy half. Portion size is the hard one — a flat photo can't see how much rice is really there, or how much oil went into the pan. So every number is an adjustable estimate, not a measurement, and I'd rather say that plainly than quote an accuracy percentage I can't back up. A real example from last week: a tester's plate came back with the jackfruit curry labelled "dhal". Both are yellow, both swim in coconut milk. Fixing that class of mistake is most of the actual work. Two decisions I made early: no ads, ever — your meal photos are never used for advertising — and no invented statistics in our marketing. Where it stands: Android and web, no iOS yet, and no barcode scanning. It reads plates, not packages. I'd love your feedback, especially on what it gets wrong. Send me a meal photo it misreads and I'll dig into it — that's genuinely how the Sri Lankan dish recognition got built.