Color blindness simulator

Normal colour vision uses three cone types, sensitive to roughly long, medium and short wavelengths. Colour is worked out from the ratio between their responses. Remove one, or shift its sensitivity toward another, and whole regions of colour space that used to produce different ratios now produce the same one.

So the effect is more specific than colours looking duller: particular pairs become the same colour. Under deuteranopia, a red and a green of similar lightness become one colour, rather than two colours that are slightly harder to tell apart.

Deuteranomaly (reduced rather than absent green sensitivity) is by far the most common form, and it is a spectrum. Someone with a mild case may never have been diagnosed and may simply find certain interfaces inexplicably annoying.

Any pair that differs mainly in hue while sharing a similar lightness is a candidate. In practice the repeat offenders are:

• Red and green, which almost every status indicator uses.
• Green and brown, common in charts and maps.
• Blue and purple, as in link states and selected states.
• Light green and yellow, as in "warning" versus "success" badges.
• Mid grey and mid anything, as in disabled versus enabled.

The tool above reports these as ΔE distances: how far apart two colours are perceptually before the simulation, and after. A pair that starts at ΔE 40 and ends at ΔE 3 has effectively merged.

It also tells you whether the pair still differs in lightness. That distinction matters more than the colours: lightness survives every deficiency, so a red and a green that differ enough in lightness usually remain usable even when their hues collapse.

The instinct is to change the colours. Usually the better fix is to stop making colour carry the meaning on its own.

Add a redundant cue. Put a tick and a cross alongside green and red, the word "Error" next to the red border, a dashed line as well as a different colour in a chart, or a pattern fill on the bar. Once the meaning survives without hue, the colour choice becomes decoration rather than information.

Separate by lightness, not hue. If two things must be distinguished by colour alone, give them clearly different lightness values. Check the pair in the contrast checker. A ratio of 1.6 or more between the two means the difference survives when hue does not.

Do not rely on a colour-blind-safe palette alone. They help, but a palette that is distinguishable in isolation can still fail at small sizes, on thin strokes, or against a coloured background.

A simulation models cone response. It does not model perception, adaptation, context, or a lifetime of learning to read colour differently. Someone with deuteranomaly is often far better at distinguishing certain shades than this simulation implies, because they have spent their life attending to the cues that remain.

So do not use this to decide "how it looks to them". Use it for the narrow, checkable thing it is good at: finding pairs that carry meaning and stop being different. That failure is real and reproducible, so it is worth fixing.