Field guide · using a photo

Show the whole dish in decent light.

A photo lookup can only work with what the frame shows. Light, framing and one familiar object for scale give it the clearest possible view.

Read with sources
Illustrative still life of a plated dish on a wooden table near a window, in soft daylight
Illustrative artwork

You do not need special equipment for a useful food photo — a phone camera and a few small habits are enough. What a lookup can see is bounded by the frame: clearer input means a clearer likely match. These tips mirror how photo-based food assessment works in research, where image quality is treated as part of accuracy.

Show the whole dish

Frame the shot so the whole plate or bowl is visible — rim, garnish and all. A cropped edge can hide exactly the clues that narrow a match: the wedge of lime at the side, the bread underneath, the sauce pooling on the rim. A little table showing around the plate is fine.

Step back far enough that nothing touches the frame's edge, then let the dish fill most of the picture. When the table holds several dishes, photograph each one separately instead of hoping a wide shot covers them all. And for dishes where the telling detail sits inside — a filled pastry, a layered bowl — a spoon lifting part of it into view shows more than untouched perfection.

Research that uses photographs for dietary assessment treats a complete, usable view of the food as a basic requirement of the method — what falls outside the frame cannot be assessed. Source: PMC review of image-assisted dietary assessment (opens in a new tab).

The same plated dish shown twice: once dark, tilted and cropped at the edge of the frame, once complete and evenly lit
One dish, two very different frames. A dark, cropped photo gives a lookup far less to work with than a lit, complete view of the same plate. Illustrative artwork.

Light beats filters

Daylight is the easiest upgrade. A table near a window, or open shade outdoors, gives the most honest colour and texture. Overhead restaurant lighting and direct flash both flatten a dish — flash especially, throwing hard highlights off sauces and glazes while darkening everything behind them. Watch for your own shadow falling across the plate, too.

Skip filters and heavy edits. Colour is a real clue: the green of a herb sauce, the char on a flatbread, the gloss of a broth. A filtered or darkened photo can hide exactly what makes a dish recognisable. If the light is dim, move the plate nearer the window rather than brightening the shot afterwards.

Reviews of image-based dietary assessment note that the method's accuracy depends on usable images, whether a trained rater or an automated system does the reading. Source: PMC review of image-assisted dietary assessment (opens in a new tab).

One dish per frame, close but complete

One dish, one photo. A crowded table forces a lookup to guess which plate you meant, and every dish ends up smaller and less readable in the frame.

Get close enough that the dish fills most of the picture while staying fully visible. Overhead works for flat dishes — pizza, a mezze spread, a noodle bowl seen from above. A slight angle suits anything stacked or layered, because the side view shows what the top view hides. Pick whichever angle reveals more of the dish.

In a systematic review comparing automated image-based dietary estimates with weighed food, errors ran lower when photos showed a single, simple food rather than a busy scene — simple framing helps people and models alike. Source: PMC systematic review of AI dietary assessment (opens in a new tab).

When the meal arrives as several dishes — a curry with sides, a spread of small plates — decide which one you want identified and give it its own frame. Photographing the whole spread as a record is fine, but the photo you submit should isolate the dish: a lookup reads the one thing in front of it far better than a table full of options.

Add a familiar object for scale

A fork laid beside the plate, a coin, your hand at the edge of the frame — anything with a known, everyday size helps a reader judge how much food there is. Image-assisted dietary studies place reference objects in the frame for exactly this reason: scale turns a shape into a serving. Source: PMC review of image-assisted dietary assessment (opens in a new tab).

Place the object beside the food rather than on it, and keep the dish itself the subject. If you would rather skip the object, a short portion note does the same job — the lookup accepts a weight, a household measure or a number of pieces. How to describe a portion without scales shows the wording that works.

An overhead view of a well-framed food photo: the whole dish centred, daylight from one side, and a fork placed beside the plate for scale
The frame, at a glance. Whole dish visible, daylight from one side, one familiar object beside the plate. Illustrative artwork.

When the clue is text — a menu line, a market board, a package label — shoot it square-on. Hold the phone parallel to the surface, close enough that the words fill the frame, and check the shot is sharp before you pocket it. A slanted, blurry label reads no better to a lookup than to you. Laminated menus and glossy packaging add one more wrinkle: tilt a degree or two to move the glare, not the camera.

For packaged food, the label itself carries verified information no photo needs to guess: the ingredients list, the allergen emphasis, the date marks. Reading it directly — the ingredients-list guide shows how — beats any lookup. The photo helps with everything that has no label.

And remember the name fallback: if the dish has a name — the menu's name, the name you heard — typing it alongside the photo gives the lookup two clues instead of one.

What not to bother photographing

A half-eaten plate gives a lookup less to work with than a fresh serving — half the clues have already been eaten. The same goes for steam-fogged lenses, blurry motion shots, extreme close-ups of texture with no context, and the whole table at once. Night-time shots under coloured lighting deserve the same caution: red neon turns every dish the same colour.

If the first shot misses, take another rather than submitting the weak one. A retake costs a few seconds; a confused match costs more trust than it is worth. Two angles cost little more — overhead for the shape, a low side shot for the height of anything stacked or layered.

And shoot before the first bite, not after it: an untouched plate shows the serving exactly as it was served, which is also the portion you'll want to describe.

When a photo still won't settle it

Some dishes genuinely look alike — flatbreads, dumplings and stews blur into each other across cuisines. A good photo earns a likely match, not a certificate. The photo-and-portion guide covers what no picture can tell you: exact weight, exact recipe, storage history. And if the photo route stalls, typing the dish name as you heard it gives the lookup a second clue.

One boundary is absolute: no photograph confirms allergens or whether a dish suits your diet. If that is your question, ask the people making the food — the allergen questions guide and the 14-allergen list walk through it.

Everything here adds up to one habit: give the lookup the clearest, simplest, best-lit version of the dish you can manage in ten seconds. The match gets better; the limits stay the same.

Have a dish in front of you? Try the lookup with these tips, and treat the answer as a starting point.

Spotted an error? Tell us — please include the page address and the line that looks wrong.