Guide

Which paints do you mix to match a colour in a photo?

An eyedropper tells you the colour. It does not tell you which of your tubes make it, in what ratio, or which one will take over the mixture.

You hold a colour picker over a photograph and it gives you #8A5A42. Now what? You do not own a tube of #8A5A42. You own eleven tubes, and the question is which of them, and in what proportion, gets you there. That gap between reading a colour and mixing it is where most of the frustration lives.

Why the screen number does not transfer

A screen makes colour by adding light. Paint makes colour by taking it away: every pigment absorbs part of the spectrum and reflects the rest, so mixing two paints subtracts twice. That is why yellow and blue give you green on the palette but a dull grey on a badly written computer model, and why every mixture loses a little value compared to the two colours that went into it.

It also means the same hex can be reached several different ways, and they will not behave the same. A brown mixed from burnt sienna and ultramarine will shift differently under a glaze, and dry to a different surface, than the same brown mixed from black and orange.

The three things that actually decide the mix

1. Which pigments you own

A recipe that calls for quinacridone magenta is useless if there is no magenta on your palette. Any answer worth having is expressed in your own tubes, which is why generic colour-matching charts fail in practice.

2. Tinting strength

Paints are not equal partners. A speck of phthalo blue will dominate a mixture that a knifeful of yellow ochre would barely move. Two paints at "50 percent each" is a statement about volume, not about influence, and this is the single most common reason a mix that should have worked comes out wrong.

3. What the light is doing

A colour in shadow is not the local colour made darker. It is usually cooler, less saturated, and it carries bounce from whatever is near it. If you sample one point and mix it in isolation, you get a passage that is technically correct and looks pasted on.

How to do it by hand

1. Mix a pile that is close, not exact. Get the value right first, then the temperature, then worry about chroma. In that order, always.
2. Hold it against the reference. Put a dab on a palette knife or a scrap of primed card and physically place it next to the colour on the print or screen. Your eye compares far better than it estimates.
3. Squint. If the two merge when you squint, the value is right. If one jumps forward, it is not.
4. Adjust one axis at a time. Too light, add the dark. Too warm, add the cool. Changing two things at once means you learn nothing from the result.

Mix from one parent, not from scratch each time

Here is the thing that fixes more paintings than any single technique. When you mix each new passage from scratch, the pieces do not belong to each other, and a face made that way falls apart even though no individual colour is wrong.

What painters actually do at the palette is mix one mother colour for the midtone, divide it into piles, and push each pile: a little white into the light, a little blue into the shadow, a touch of rose where the ear turns warm. One mix carries the whole passage, and that shared parent is what makes skin read as skin instead of as a patchwork.

So when you are working from a reference, get your midtone right first and treat everything else as that midtone plus one adjustment. It is faster, and it holds together.

How BrushLens does it

The app compares a sampled colour against every paint on your palette using CIEDE2000, the perceptual colour difference standard, rather than raw RGB distance, because two colours can be numerically close and look nothing alike. It then searches pairs and triples of your paints, modelling the mixture as a weighted geometric mean in linear light, which is what makes yellow plus blue come out green and makes mixtures darken slightly the way real paint does.

It will not add a third pigment unless the third genuinely improves the match, because a three-pigment mix is harder to repeat and quicker to go muddy. And it can hold a base mix fixed at 50 percent or more, so every reading comes back as your base plus what you add rather than a fresh recipe each time.

Written by Soren Ulsig. Last updated 2026-08-29.