AI photo calorie counter & scanner

Manual calorie tracking is the most common reason people quit logging. Searching a database for "grilled chicken," picking the right entry from 47 options, entering a portion estimate — that's 2-4 minutes per meal, three times a day, every day. AI photo logging compresses that to about 15 seconds. Here's how it works in Makrosas.

How AI photo logging actually works

1. You take a photo of your plate. From above is best, with a fork or known object for scale.
2. AI identifies each food item. "This is chicken breast, this is rice, this is broccoli."
3. AI estimates portion size. Volume detection from the 2D image, calibrated against known references.
4. Calories and macros appear. Already logged to your day. Tap to edit if anything looks off.

The whole flow is roughly 3-4 seconds for the AI to process, plus your time to take the photo. End-to-end ~15 seconds per meal.

What Makrosas's photo counter actually recognizes well

• Common single foods: eggs, fruit, single piece of meat or fish — 95%+ accuracy.
• Standard composed plates: protein + grain + vegetable combinations — 88-92% accuracy.
• Packaged products with visible labels: 100% — the AI reads the nutrition panel directly.
• Restaurant-style plates: works for most Western and increasingly for regional cuisines (we add training data for under-represented foods continuously).

Where the AI struggles, and what to do

Hidden oils and sauces

Olive oil in a salad, butter in a stir-fry, dressing on a bowl — these are invisible to the camera and the #1 source of error. Fix: add free-text after the photo: "with 1 tbsp olive oil." The AI updates calories accordingly.

Mixed and layered dishes

Soup, stew, casserole — the camera only sees the top. Fix: describe what's underneath in chat: "beef stew, 1.5 cups, with potatoes and carrots."

Plate-size illusions

A small plate looks full. A large plate looks modest. Fix: include a fork or standard utensil in frame — gives the AI a fixed scale reference.

Why this matters more than you think

The decisive factor in tracking success is consistency, not precision. Studies repeatedly show that people who log 7 days a week — even with imperfect data — outperform those who log perfectly for 3 days then quit. AI photo logging removes the friction that causes the quit.

Five minutes saved per meal × three meals × seven days = nearly two hours saved per week. That's two hours of not hating the tracker, two hours of staying engaged, two hours that compound into 90-day habits instead of 14-day attempts.

How to get started

1. Download Makrosas (iOS or Android, both free).
2. Skip onboarding shortcuts — tell the AI about you in plain English ("I'm trying to lose 8 kg, I sit at a desk, I train 3x/week").
3. Open the camera, photograph your next meal, accept the result. That's it.
4. For a week, log every meal by photo. Notice the difference vs. your old tracker.

Frequently asked

Is the photo logger really free?

Yes. AI photo recognition is in the free tier — no paywall, no daily limit beyond reasonable use. (Premium, planned later, will add advanced analytics and unlimited AI chat context — not photo logging itself.)

Can I correct a wrong AI result?

Yes — every photo result is editable. Tap the food item, change the portion or item type, save. The AI also accepts free-text refinements ("actually that was sweet potato, not regular").

What if my food isn't in any database?

Free-text input works as a fallback: "ate two homemade buckwheat pancakes with cottage cheese, about 200g total." The AI computes macros from the description even without a database match.

Does it work on European foods?

Yes. The model is trained on a broad European cuisine set including Mediterranean, Western European, Central European, and regional Eastern European dishes. Common items like schnitzel, goulash, paella, risotto, blini are recognized.