Eat. Log. Don't stress.
The fastest way to log a meal is the way you'd describe it to a friend. So we built the app around that.
Take a photo of your plate.
Or just say what you ate.
Or type. 14 words or 4.
A coach who knows your week.
iOS and Web. Same account.
A few habits make it sharper.
Logging is one tap, but the estimate gets better when the photo gives the AI more to work with. These are the tricks our most accurate users rely on, starting with the one that helps most: weigh it and let the AI read the scale.
- 01Angle
Hold the phone at about 45 degrees
A slight tilt instead of a straight-down shot lets the AI see the height of the food, not just its outline, so it can judge volume on bowls, stacked sandwiches, and piled rice. For a flat plate where nothing is stacked, a clean top-down shot works just as well.
- 02Framing
Leave nothing hidden
Photograph the meal before you start eating, with the whole plate in frame and items not overlapping. Food tucked under a bun, buried in a bowl, or stacked behind something else cannot be measured. Spread the components out a little if you can.
- 03Weight
Put it on the scale and shoot the readout
Set the plate or bowl on a kitchen scale and include the number in the photo. The AI reads the grams off the display and uses the exact weight instead of estimating volume from pixels. This is the single biggest accuracy upgrade for anything you portion at home.
- 04Detail
Say what the photo cannot show
Add a line of text for anything invisible: cooked in two tablespoons of olive oil, low-fat milk, no dressing, restaurant portion. Oils, butter, sauces, and the sugar in your coffee rarely show up in a picture and can quietly add 100 to 400 calories.
- 05Scale
Add a size reference
No scale handy? Put a familiar object in the shot, a fork, a spoon, a coin, or a credit card, and the AI uses its known size to scale the portion. Your hand resting next to the plate works in a pinch too.
- 06Detail
Name the restaurant or brand
For chain food, type the place: Chipotle chicken bowl, Starbucks grande latte, a Big Mac. The AI pulls the exact published nutrition for the named item instead of guessing from the picture, which makes branded meals close to exact.
- 07Light
Use good, even light
Natural daylight gives the truest colors and the sharpest detail, and color and texture are exactly what the AI uses to tell foods apart. Avoid dim restaurant lighting, hard shadows, and strong color casts from screens or neon signs.
- 08Quality
Keep it sharp and fill the frame
Hold steady, let the camera focus, and get close enough that the meal fills most of the frame without cropping anything off the edge. A blurry or distant photo hides the texture cues the model relies on to identify each food.
- 09Framing
Shoot multi-part meals piece by piece
For a homemade plate with several distinct foods, a quick photo of each part beats one busy shot. Separate photos give the model clean boundaries for the rice, the chicken, and the vegetables instead of one blurred pile.
- 10Review
Check the portion, then nudge it
The number the AI returns is a starting point, not the final word. If a portion looks light or heavy, drag the one-tap slider and the totals update live. A two-second correction often does more for accuracy than a perfect photo.