How to Log Macros with AI Photo Logging (Tips for Accurate Results)

Photo Logging Is the Fastest Way In — When You Do It Right

Typing what you ate is quick. Snapping a photo is quicker still — you point your camera at the plate, and the macros come back. But “point and shoot” hides a real skill: how you take the photo decides how good the estimate is.

Photograph food thoughtlessly and the AI gives you its best guess from what it can see. Photograph it deliberately — the right angle, the packet in frame, a sense of portion — and the estimate tightens dramatically. This is a short guide to getting the most out of AI food logging with your camera, plus the one feature that makes photo logging pay off for months: NomsAI remembers every item you shoot.

How Photo Logging Actually Works

When you take a photo in NomsAI, the image is sent to a vision AI model that identifies the food and estimates its calories, protein, carbs and fat. It reads what’s on the plate — the type of food, roughly how much, how it looks like it was cooked — and returns a macro breakdown you can review and tweak before saving.

It’s the same engine as describing your meal in text, just fed by a picture instead of a sentence. And like any estimate, it’s only as good as the information you give it. A blurry shot of a beige plate is a hard problem. A clear shot with a brand label in frame is an easy one.

The Single Best Habit: Photograph the Packet

Here’s the tip most people miss. When a food has packaging, put the packaging in the shot.

A photo of a protein bar is just “a protein bar” — the AI estimates from the category. A photo of the wrapper is a specific product: it can read the brand, the flavour, often the exact nutrition panel. The difference in accuracy is enormous, because you’ve handed it a real label instead of asking it to guess.

This turns your camera into something better than a barcode scanner. Barcodes only work if that exact product is in a database. A photo of the packet works even for a new flavour, a regional product, or something that never made it into any database — the AI reads the pack the way you would.

So for anything packaged — ready meals, snacks, sauces, branded drinks, a bag of a specific supermarket’s granola — get the front of the pack, and ideally the nutrition label, into the frame. It’s the highest-leverage thing you can do.

Log It Once, Then Type It Instead

Photo logging isn’t only about the meal in front of you — it teaches NomsAI your foods.

Every item you log is remembered on your device, so the next time you start typing that food it appears as a suggestion. You pick it instead of describing it — or re-photographing it — from scratch. Shoot your usual brand of yoghurt once with the label showing, and NomsAI learns that exact item; from then on you just type “yog” and select it. No hunting for the packet, no retyping the full brand.

It still logs the normal way — selecting a remembered item runs it through the AI just like a fresh entry, so you’re not skipping the estimate or the fuel it uses. What you’re skipping is the effort: a fiddly branded product becomes a two-letter shortcut. That’s exactly why the first capture is worth doing well. Get the specific product right once, with the packet in frame, and it’s quick to find and re-log every time after.

Tips for a Better Estimate

A few habits noticeably improve what comes back:

• Fill the frame with the food. Get close. A plate lost in a wide shot of the table gives the AI less to work with.
• Shoot from a slight angle, not straight down. A three-quarter angle shows depth and volume — a burger’s height, how full a bowl is — which helps portion estimation more than a flat top-down view.
• Get the packet or label in frame for anything branded (see above). This is the big one.
• Give it decent light. Natural light beats a dim restaurant. The clearer the image, the fewer assumptions the model has to make.
• One food per photo when portions matter. A single item is easier to size accurately than a crowded plate. For a mixed plate, a plain-text description (“half a roast dinner — chicken, roast potatoes, veg, gravy”) sometimes lands better than a photo.
• Always glance at the review screen. The AI shows you what it found before you save. If a portion looks off, nudge it. Ten seconds of sanity-checking keeps your day honest.

Where Photo Logging Shines — and Where It Doesn’t

Being straight about this matters more than hype.

It shines at: packaged foods with a visible label (near-exact), single clear items, and anything you can’t easily put into words — a mixed salad, a colourful bowl, a pastry you don’t know the name of. Point, shoot, done.

It struggles with: deeply mixed dishes where the calories hide (a creamy curry, a casserole — you can’t photograph the butter and oil stirred in), foods that look identical but aren’t (full-fat vs low-fat, sugar vs sugar-free), and portion size from a bad angle. In those cases the AI is guessing at what it genuinely can’t see — and that’s not a camera problem, it’s a physics problem. No tool reads hidden ingredients from a photo.

The honest rule: use the photo for what’s visible, use your words for what isn’t. If oil, sugar or a specific brand is doing the damage and the camera can’t see it, tell it in text. Often the best log is a photo plus a quick note.

It Doesn’t Have to Be Perfect

Accuracy anxiety is what makes people quit tracking. You don’t need to be within a gram — you need to be roughly right, consistently. For managing weight or understanding your diet, landing within ~100 kcal of the truth and actually logging it beats a perfect number you skipped because it felt like too much effort. A photo you took in five seconds and saved is worth far more than the meticulous entry you never made. (More on this in tracking calories without weighing your food.)

A Note on Privacy

Food photos can carry hidden metadata — where and when the shot was taken. Before your photo is processed, NomsAI strips that metadata, so no location or device information travels with it. The image is used to work out your macros and isn’t kept on our servers; a small thumbnail stays on your device so you can see what you logged. Privacy is the point of the whole app — you can read more about why that matters.

The Bottom Line

Photo logging is the fastest way to track macros — but do it deliberately. Get close, use a good angle, and above all get the packet in frame so the AI reads the exact brand instead of guessing. Confirm the estimate, and remember that the camera can only see what’s visible — describe the rest in words.

Do the first capture well and NomsAI remembers the item: after that it’s a quick type-ahead instead of a fresh photo every time — the specific product you shot, ready to re-log in a couple of keystrokes.

NomsAI does all of this with no account and no forced login — type it, say it, or snap it, get your macros, and get on with your day.