Extract colors from a photo

How the dominant colors are picked

The tool samples the pixels of a downscaled copy of your image and groups them into families of similar shades (a technique called quantization). The largest families become the palette, and near-duplicate shades are filtered out along the way. The result is 6–8 colors that genuinely represent the photo, not washed-out averages that all drift toward gray. It's the same principle design tools use to build palettes from moodboards and photography.

What you can do with a palette pulled from a photo

Starting from a photo's colors is the fastest way to keep your designs consistent: a video thumbnail that matches its cover frame, a landing page in your product's colors, a slide deck that harmonizes with the images inside it. Each HEX code (e.g. #FF8A5C) pastes straight into Canva, Figma, Photoshop, PowerPoint, or the CSS of your website.

What «dominant colour» means

It is not the colour you notice most, it is the one occupying the most pixels. Those are two different things, and the difference explains almost every palette that looks wrong: in a sunset photo the dominant colour may be the grey of the road, because the road takes up half the image, while the orange that strikes the eye sits in a strip.

The method groups pixels into families of similar tones and takes the largest families: it is called quantisation. It follows that a large uniform background nearly always wins, and that cropping the photo to the part you care about changes the result completely.

How an extracted palette is used

The most useful rule in graphic design is 60-30-10: one colour dominates the surface, one accompanies it, a third appears only in accents — a button, a heading, a detail. Taking four colours from a photo and using them in equal parts nearly always produces a muddle.

The second criterion is text contrast: a colour that looks lovely on screen is not necessarily legible under lettering. Dark text on a light ground, or the reverse, has to be checked, and for accessibility the recommended contrast ratio is at least 4.5 to 1 for normal text. Many natural palettes are all mid-tones, and in that case a very dark or very light colour has to be brought in from outside.

Why two tools give different palettes

Because the word «similar» has no single definition. Some group colours in RGB space, that is by the numbers, others in spaces built on perception, where two nearby colours are the ones the eye judges to be close. The latter give more convincing results on skin tones and greens, where RGB errs badly.

Then there is the number of families requested and the fact that the image is reduced first before being analysed, something everyone does for speed and which blends neighbouring colours. So there is no «right» palette for a photo: there are plausible palettes, and it is best to pick one and then adjust it by hand.

The three traps of colour taken from a photo

The colour profile: a photo in Adobe RGB read as sRGB gives wrong hex codes, and the colours come out dull. Compression: in a heavily compressed JPEG the colours near edges are invented by the algorithm, and picking a colour exactly there returns a hue that was never in the scene.

And print: screen colours are light, print colours are ink, and very saturated oranges and greens do not exist in print. A palette chosen on screen for a business card should always be tested on paper before ordering a thousand copies.

Nearby tools

To work on colours there are Colour converter and Palette generator. On the source image Black and white photo and Resize photo, and for a background Blur part of a photo.