Colorblind Palette Checker
Here is the question a contrast checker cannot answer: two colours can each pass against the background and still be the same colour as each other to someone with colour vision deficiency. Paste your palette and see which pairs collide — and by how much.
Your palette
Or paste CSS, Tailwind config, JSON, or a list of hex codes
Red-green CVD (~1 in 12 men)
Collision2 collisions · 1 risky · 6 pairs
Blue-yellow CVD (~1 in 10,000)
Risky
Red-green and blue-yellow are scored separately on purpose. Averaging them marks palettes like Okabe-Ito — designed specifically against red-green CVD, where it scores zero collisions — as failing on the strength of one close pair under a condition affecting ~1 in 10,000. A rare type should not overrule the common one, but it should not be hidden either.
This is a measurement, not a certification
Simulation uses Machado 2009 for red-green and Brettel 1997 for blue-yellow; separation is measured as CIEDE2000 in CIE Lab. The algorithms are named because many online tools do not say what they use. But colour vision varies between individuals with the same diagnosis — two protanopes can disagree about the same pair — so what this page gives you is a checkable number and a flagged risk, not a certificate of safety. For work that matters, show it to real users.
Pairwise collision matrix
Click any cell to see what the two colours actually look like with colour vision deficiency.
| — | ||||
| — | ||||
| — | ||||
| — |
Each cell shows ΔE2000 (worst across red-green CVD). Below 10 is a collision; below 20 is risky.
Share this check
The link carries the palette, so whoever opens it sees exactly this result.
Export
:root {
--color-1: #1f77b4;
--color-2: #ff7f0e;
--color-3: #2ca02c;
--color-4: #d62728;
}Load a reference colourblind-safe palette
Each is listed with its primary source, author and licence. Click to load and check it.
Okabe-Ito
Masataka Okabe (Jikei Medical School) & Kei Ito (University of Tokyo) · 2002, revised 2008
The de facto standard for colourblind-safe scientific figures. Both authors are strong protanopes and designed it from lived experience rather than theory — which is why it holds up: zero red-green collisions across all 28 pairs.
Primary source · No formal licence. The authors ask only to be credited: “Please feel free to use these items for your classes and seminars. (Please don’t forget, however, to mention Masataka Okabe and Kei Ito for reference.)”
Paul Tol — Bright
Paul Tol (SRON Netherlands Institute for Space Research) · 2021 (technical note issue 3.2)
Seven saturated colours for line plots and categorical data, designed against red-green CVD. The grey is reserved for 'bad' or missing data, so it carries meaning — don’t reassign it to a category.
Primary source · BSD 3-Clause (declared in the author’s tol_colors.py)
Paul Tol — Muted
Paul Tol (SRON Netherlands Institute for Space Research) · 2021 (technical note issue 3.2)
Nine lower-saturation colours — the largest qualitative set here that still clears red-green CVD cleanly. Useful when seven is not enough and you would otherwise reach for an unsafe scheme.
Primary source · BSD 3-Clause (declared in the author’s tol_colors.py)
Paul Tol — High Contrast
Paul Tol (SRON Netherlands Institute for Space Research) · 2021 (technical note issue 3.2)
Three colours that survive greyscale printing as well as colour vision deficiency — the only set on this page that scores safe on every axis, including achromatopsia. Use it when the figure may be photocopied.
Primary source · BSD 3-Clause (declared in the author’s tol_colors.py)
ColorBrewer Dark2 (3 classes)
Cynthia A. Brewer & Mark Harrower, The Pennsylvania State University · 2002
Shown at three classes deliberately. ColorBrewer flags Dark2 as colourblind-safe only at three — and our own analysis agrees: three colours collide zero times, while the full eight-colour version collides six times. The scheme is safe; the eight-colour version of it is not.
Primary source · Apache-style licence. Attribution is mandatory, and the ColorBrewer name may not be used to endorse or promote derived products.
ColorBrewer Set2 (3 classes)
Cynthia A. Brewer & Mark Harrower, The Pennsylvania State University · 2002
A softer alternative to Dark2, and subject to the same three-class limit. Of ColorBrewer’s eight qualitative schemes only three are ever flagged colourblind-safe, and never beyond four classes.
Primary source · Apache-style licence. Attribution is mandatory, and the ColorBrewer name may not be used to endorse or promote derived products.
Cividis (sequential)
Jamie R. Nuñez, Christopher R. Anderton & Ryan S. Renslow (PNNL) · 2018
A continuous ramp optimised specifically for colour vision deficiency (PLOS ONE 13(7): e0199239), so that someone with deuteranomaly reads it nearly the same way as someone with typical colour vision. Unlike viridis — whose authors claim only that it is ‘relatively friendly’ to colour blindness — cividis was designed for it.
Primary source · CC0 public domain
Viridis (sequential)
Nathaniel J. Smith, Stefan van der Walt & Eric Firing · 2015
Perceptually uniform and monotonically increasing in lightness, which is why it survives greyscale printing. Its authors describe it as ‘relatively friendly to common forms of colorblindness’ — a weaker claim than ‘colourblind-safe’, and worth repeating accurately. For a CVD-optimised ramp, use cividis.
Primary source · CC0 public domain
Attribution
This product includes color specifications and designs developed by Cynthia Brewer (http://colorbrewer.org/).