Log a Meal From a Photo: What the AI Sees and What It Needs From You
Quick answer: To log food with a photo, photograph the whole plate before you eat, let the AI identify the dish and estimate the portion, check the weight in grams it shows and correct anything you know is off, then add what the camera could not see by voice or text. The photo can't see the things that make food heavier than it looks: cooking oil, butter, dressing, sugar stirred into a drink, and anything under the top layer of a bowl. Those few words from you matter more than a better photo.
Step by step
- Photograph the whole plate before the first bite. One frame, everything in it. A half-eaten plate is harder to read and you will not remember what was on it.
- Let the AI read it. In CalMePlease the model identifies the dish and counts calories and macros. Photo analysis runs on the server, so it needs an internet connection.
- Check the grams. The app shows the estimated weight in grams. If you know the rice was a smaller scoop or the steak was bigger, correct it; this is a one-tap change.
- Add what the camera missed. Oil in the pan, the dressing, the sugar in your coffee: log them by voice ("a tablespoon of olive oil") or as text. Voice and text work in nine languages.
- Several dishes on the table? Record a short video instead. A short clip lets the model see each dish from several angles, and each dish lands as a separate entry with its own weight, so fixing one does not change the others.
- Forgot to log it? Pick the past day in the calendar and add the meal there; daily totals and the weekly report recalculate.
What the photo sees and what it doesn't
A photo is good at what is on the plate and rough about how much. It cannot see inside the food or under it. The table below is the checklist: if your meal has anything in the right-hand column, add it in words.
| The photo sees | The photo doesn't see | What to add by voice or text |
|---|---|---|
| A salad and its leaves, tomatoes, cheese | Dressing already mixed in, oil on the leaves | "With about two tablespoons of Caesar dressing" |
| Fried eggs, a stir-fry, roasted vegetables | Oil or butter used for cooking | "Fried in a tablespoon of butter" |
| A cup of coffee or tea | Sugar, syrup, the kind of milk | "Two sugars, whole milk" |
| The top of a bowl of rice, pasta, soup or porridge | How deep the bowl is and what is underneath | "Big bowl, about two cups of rice" |
| A sauce on top | Cream, sugar or oil inside the sauce | "Creamy sauce" or "restaurant curry, quite oily" |
| A glass of juice or a smoothie | What went into it, sweetened or not | "Orange juice, no added sugar" |
| A sandwich or a burrito from the outside | Fillings, mayo, the number of slices of cheese | "Chicken, mayo, two slices of cheese" |
| Yogurt, milk, cheese | Full-fat or low-fat | "Full-fat Greek yogurt" |
Why this pattern exists, and how the four stages of recognition work, is explained in How AI food recognition works. How large the error is with and without a description is covered in How accurate is AI calorie counting by photo.
Getting a better estimate
The model can only use what is in the frame. These habits put more of the meal into it:
- Shoot the whole plate, before eating. Edges of the plate in the frame give the model a sense of scale. A cropped close-up of one item loses it.
- Use an angle of about 45 degrees. Straight down hides height: a flat layer of rice and a heaped one look the same from above. Side-on hides what is behind the front item. An angled shot shows both the surface and some depth.
- Use daylight or even indoor light. Dim light, a strong colour cast or a hard shadow across the plate make foods harder to tell apart.
- Leave a familiar object in the frame. A fork, a standard mug or your hand next to the plate gives a size reference. Do not stage it; just do not move it out of the way.
- Separate what you can. Food side by side is easier to read than food stacked. If the sauce comes in a dish, photograph it before pouring.
- One plate, one photo. For a shared table with several dishes, a short video pan works better than one wide photo.
- Correct, do not re-shoot. If the dish is right and the portion is wrong, fixing the grams is faster and more accurate than taking a better picture.
When to use voice, barcode or text instead
A photo is the right tool when the food is in front of you and you are not sure exactly what is in it. It is the wrong tool in a few common cases:
- Packaged food: scan the barcode. The numbers come from the label, not an estimate. Then say how much of the package you ate. See Barcode scanner for calories: what the label doesn't tell you.
- Snacks and drinks, or food you already ate: use voice. A handful of nuts or a flat white is faster to say than to photograph, and you can log a meal after the plate is gone. See Voice food logging.
- A recipe you know by heart: type it. If you cooked it and know the ingredients, type the dish in words; the AI breaks it into ingredients and fills in the grams.
- No internet: use voice, text or barcode. These are stored locally and sync when you are back online; photo analysis waits for a connection.
FAQ
Do I have to photograph every meal? No. Photo is one of four ways to add a meal in CalMePlease, alongside voice, barcode and text (video sits next to photo). Use whichever is closest to hand for that meal.
Can I fix the portion after the photo is logged? Yes. The app shows the estimated weight in grams, and you can correct it in one tap. Corrections feed back into the next estimate.
Does photo logging work offline? Not the recognition itself: it runs on the server and needs a connection. Text, voice notes and barcode scans are stored locally and sync later.
What about a table with several dishes? Record a short video pan instead of a single photo. The model reads several dishes in one clip and logs each as a separate entry with its own weight.
Is photo logging free? The app is free to download and you can log meals right away. A subscription unlocks unlimited photo recognition, the AI coach and full statistics; prices are listed on the App Store page.
Can I log a meal I forgot to photograph? Yes. Choose the past day in the calendar and add the meal by voice or text; totals and the weekly report update.
Data: Product features as described on calmeplease.com and the App Store page, checked October 2026.
Related: How AI food recognition works · How accurate is AI calorie counting by photo · Voice food logging
CalMePlease logs a meal from a photo or a short video, then lets you add the oil and the sugar by voice or text in nine languages. Free on the App Store.
