https://www.inclusivecolors.com/
So one approach is you create swatches of different colors that go from grade 100 (light) to grade 900 (dark), where the lightnesses are chosen such that all grade 700 colors contrast against grade 100 colors, all grade 800 colors contrast against grade 200 etc.
And then you know red-700 vs gray-100, green-800 vs yellow-200 and so on will contrast without having to check.
If you go to the Contrast menu, you can also explore how much stricter the APCA algorithm (meant to be more accurate) is compared to WCAG. For dark on light colors especially, APCA is much stricter about what contrast so you really shouldn't use WCAG for dark themes.
Also, if you go to the Examples menu and check out the Tailwind and IBM Carbon color palettes, you can see how the swatches in hand designed palettes vary their saturation and hue across grades in a non-linear way. So automatically picking if white/black contrasts the best is more straightforward (like the article mentions), but for more deliberate/branded palettes, you can't just generate a color with a simple lightness component shift, so this is more open ended.
--text: lch(from var(--bg) calc((49.44 - l) * infinity) 0 0);
https://evilmartians.com/chronicles/oklch-in-css-why-quit-rg...
Can honestly say this article changed my perspective on this subject drastically, such an amazing tool. I was very surprised that my designer friends hadn't heard of oklch at all, it solves a whole category of problems.
For those making anything at a production scale where you need wcag compliance however, I'd avoid this and leverage a proper semantic token layer. Semantic tokens will help both accelerate your dev cycle, and they'll help guarantee proper contrast ratios in a way that looks visually better than just switching your foreground layer to black or white. The great thing about a semantic token layer is they're extremely easy to theme, which means you get light/dark theming for very little additional cost. You can also create separate WCAG2 / APCA accessible themes, should your brand color be one of the ones that WCAG2 has issues with - will get you compliance while still providing a better visual contrast option.
This is kind of my niche domain specialty - I run the variables/tokens stream at Figma, and I've worked on the dark mode implentation for both Figma and Atlassian. Happy to answer any questions about tokens/themes/accessible color.
This exact type of functionality has caused a major project a work on to use CSS in JS (for relative colors and contrast colors.
I’m glad to see this type of thing coming around the corner and look forward to it being widely available in a couple years.
* bg-brand (this would be used whenever you need your brand color as a background)
* text-danger (likely a red text color)
* icon-warning-hover (likely a dark yellow-orange that's slightly different from icon-warning)
Generally speaking, there are three "levels" of tokens: primitive, semantic, and component. Primitive tokens describe the value. In the case of color, this might be a color ramp. IE red/100, red/200, red/300. Semantic tokens reference primitive tokens. IE bg-brand might have its value set to blue/300. This layer is sometimes called a "reference" layer because of this, but I'm not a fan of that nomenclature since the component layer also references the semantic layer. The component layer is one that describes where in a component the token should be used, ie button-bg or button-text. I highly, HIGHLY recommend against using a component layer though in all but the most extreme multi-brand situation. If you aren't unilever, you should never use component tokens.
https://medium.com/eightshapes-llc/naming-tokens-in-design-s...
The primitive/semantic/component set of tiers are a general practice. Naming within them heavily differs (and should!). The names you use for the individual tokens depend on goals and intent - ie Google’s material’s semantic layer uses a naming schema that’s designed for colorful variety of themes (albeit at the expense of clarity of how they should be used), whereas Apple uses a far more simplified naming schema since the design of their apps has far fewer design differences.
As your link covers, there’s then a million different ways to implement/extend that based on whatever theming and systems you’re implementing on top.
I like the feature but in a corporate site/application, you don’t want to rely on this function because you cannot control what the result is going to be. For all I know, WebKit could fix some later bug or change something that changes the result color to something that I don’t want.
Cant someone take a look at the buttons before the large project ships? Alternatively make it mandatory to never have black text on a dark button and tell every team member including the large ones.
Interesting to read about the perceptual contrast vs mathematical - I did not know that. Going to integrate that into my workflow.
They can, of course, but this is how you end up with pre-release regression testing cycles that are weeks or even months long. A "large project" might easily have thousands of buttons (or more!), many of which are only seen when certain settings are enabled, certain options are chosen during a complex workflow, etc.
> APCA says you don't need as much contrast
You can always specify the threshold if you want, e.g. "apcaContrast(color)) >= $targetContrast" after adjusting, depending on what you want to do.
It really is easy, just make sure you have enough color space.
re: just change APCA contrast target, that's separate from the Not Even Wrong stuff in the article. I didn't mean to imply APCA is wrong to say you need less contrast, but rather, that the article is wrong to conclude white has more contrast.
You are right, 0.18 is indeed perceptually closer to "middle gray" because the eye responds more sensitively to darker tones, so yeah, using a threshold closer to 0.18 makes more sense if we want to identify whether a color "feels" light or dark.
That said, 0.5 is a mathematical midpoint, but as I said, not aligned with how humans perceive brightness (edit: it is aligned).
Ultimately one could use 0.18-0.3 as threshold.
There are two physical quantities for luminance, relative, and perceptual, so that passed along a nugget for those not as wise as you who might not know that :) As you know and have mentioned, using 0.5 with the luminance calculation you mentioned, for relative luminance, would be in error (I hate being pedantic, but it's important for some parties, a11y is a de facto legal requirement for a lot of work, and 0.5 would be spot on for ensuring WCAG 2 text contrast as long as used with perceptual luminance, L*)
> doesn't align with human perception
It is 100% aligned with how humans perceive brightness, in fact, it's a stable work product dating back to the early 1900s.
> Ultimately one could use 0.18-0.3 as threshold
Perceptual luminance and relative luminance have precise mathematical definitions, one can be calculated in terms of the other.
If you need to hit contrast K with background color C, you won't be able to treat this as variable. What you pass along about it being variable is valuable, of course, in that, given K and C, output has a range, i.e. if contrast algo says you need +40 L* for your text to hit APCA/WCAG whatever, and your C has 50 L*, your palette is everything from 90 L* to 100 L* and 0 L* to 10 L*.
BTW, would this relatively simple way to determine if the color is dark work?
$luminance = 0.299 * $r + 0.587 * $g + 0.114 * $b;
return $luminance < $threshold;
No it doesn't? The screenshot of the calculator in the blog post very clearly shows that white has a greater contrast according to APCA. (If the negative numbers are confusing, you can also put the colors into a BridgePCA calculator like https://www.color-contrast.dev/?txtColor=FFFFFF&bgColor=317C... to see WCAG-2-style "contrast ratio" metrics computed using APCA.)
The point of APCA is to make the contrast calculation more perceptually accurate, not just lower the threshold.
If you're unlucky enough to be familiar with the math, it is trivial to show it lowers the threshold at luminances, no matter the luminance (widely accepted Y or L*, or APCA's own luminance calc)
Re: the idea I think the point of APCA is to lower contrast threshold - you're entirely correct that this is a side effect of the main goal - to model contrast more accurately.
There is no such thing as non-perceptual contrast, perceptual contrast is a tell stuff is being regurgitated.
We're a bit far afield when we're doing APCA => BCPA => claims actually contrast ratio is 2.4 => look at actual contrast ratio, and its...5.39.
2.4 is supposed to be unreadable, which clearly isn't the case here.
So what's going on?
The wrong text size is being used on the BCPA site, it's being calculated as if it's 12 pt and it's 36 pt in the article.
The article says the standard specifies the calculation to use.
Consistent, it is not. Ex. we can imagine a background at L* 50 that is ~equally served with a white or black foreground - in that case, the aesthetic principles come into play.
To also disambiguate that, and get to 100% reliable, if both a darker and lighter color are available given contrast K and background color C, look at C, if it's L* is >= 60, choose lighter.
Then, it is 100% correct and consistent.
https://i.sstatic.net/18bQt.png
I forget the calculation but thinking about it you can probably just take the average of the 3 rgb values and compare them(?) It would produce a low value for blue and give preference to white text.
https://www.freepik.com/free-vector/white-3d-editable-vector...
https://youtu.be/tUJvE4xfTgo?si=vFlegFA_7lzijfSR (warning: video is in Portuguese)
https://news.ycombinator.com/item?id=44015990
Video seems fine. I don't speak Portuguese though so can't judge what you said but code looks good!
What really needs a fix is when you have text over an image or other diverse background (like, sticky/fixed text over a scrolling background) and need to have it always visible. And... this doesn't help at all.
So not only does this only (maybe) help in very questionable circumstances, they needed to come up with an entirely new verb for it, it has an anemic feature set (only selects black or white), and they did it with the worst possible contrast selection algorithm (doesn't select the choice with the most perceptual contrast). Way to go!
Plenty of web sites allow the end-user to select colors[1], or automatically derive colors from assets provided by the end-user. For those that care about accessibility, they typically calculate contrasting colors to prevent the user from creating a non-accessible experience. A built-in CSS tool like this will, hopefully, encourage more sites to provide a basic amount of accessibility while in no way hindering those who want to build an even better experience.
It would be cool if this was more customizable like the npm contrast-color package but the blog post details why they started with white/black with intentions of changing the algorithm later.
[1] Example: https://coolors.co/8fbfe0-7c77b9-1d8a99-0bc9cd-14fff7
I'm surprised parent hasn't come across this usage, I see it everywhere.
They specifically say they are following WCAG 2 algorithms, and that WCAG 3 may correct this issue. They say that they can easily adjust to use the better algorithm in the future when it's standardized.
It will take some time until this feature is broadly available, and I'm having some doubt that it will be implemented in the same (or correct) way on all platforms.
https://akkartik.name/post/2025-04-04-devlog
One difference in my approach is: it's an authoring-time tool. If no sufficiently contrasting color exists you get an error. And so you have to change the background until there is one.
Safari has had several security fixes since then, so you should update:
18.3: https://support.apple.com/en-us/122074
18.3.1: https://support.apple.com/en-us/122285
18.4: https://support.apple.com/en-us/122379
18.5: https://support.apple.com/en-us/122719
Also, 18.4 was a pretty big update for standards support and other features:
https://webkit.org/blog/16574/webkit-features-in-safari-18-4...
The [draft for addition to the CSS spec](https://drafts.csswg.org/css-color-5/#resolving-contrast) just got added. Kind of wonder why Apple includes this in Safari as color-contrast rather than -webkit-color-contrast until other browsers have at least indicated a position on this draft. All I can find is decisions to defer the specifications (going back to as far as 2020).
> Support for this feature first shipped in March 2021, in Safari Technology Preview 122.
https://webkit.org/blog/11577/release-notes-for-safari-techn...
> Added experimental support for CSS Color 5 color-contrast()
"Color Wheel: The Basic Color Theory for Artists and Designers" https://dessign.net/color-wheel-theory/
- Their work does ensure contrast.
- The white on blue clearly has less contrast, not more. (squinting is a cheap way to test, or, walking backwards from your monitor)
With APCA, backgrounds around L* 60 tend to still allow white foregrounds, which is aesthetically closer to what the eye wants.
A black foreground would have more contrast regardless, even by APCA.
To be fair, this is how APCA is almost always demonstrated as a win over the long-running standard, so people run with the premise that the demo image of APCA is more contrast, rather than "ours say you'll have enough contrast to be accessible with a white foreground, even if it also says the contrast would be higher with a black foreground".
(source: in 2020 built color system around the same science, enabling latest iterations of Material theming)
Is your screen really badly miscalibrated, or do you have some unusual vision condition? That’s all I can think of. I agree with the article, the white is very clearly higher contrast.
> A black foreground would have more contrast regardless, even by APCA.
OK, now I’m just baffled. The article shows the lightness contrasts for white and black on that particular blue: black gets Lᶜ 38.7, white gets Lᶜ −70.9. White foreground has more contrast, according to APCA.
I really am baffled by what you’re saying, because it all sounds coherent… except it’s all back to front.
I can't speak to Material You, but I've seen this sort of thinking at companies that are more concerned with legal compliance with the strict wording of WCAG 2, rather than on-the-ground user experience. People can even learn to ignore their lying eyes and fairly accurately guess what the WCAG 2 "contrast" metric for a given pair of colors will be, independently of how easy or hard the colors are to distinguish from one another.
Hopefully WCAG 3 will incorporate better color guidance from places like APCA, and at the very least these companies will stop producing unreadable black-foreground buttons and badges all the time.
EDIT:
I get it, it is easily read as "the entire article is wrong" instead of "the article is wrong on these points"
You're free to elaborate on your concerns. We could raise this to a conversation, I think that'll feel better for both of us than me taking that remark about me personally.
For example, I agree that the primary container color shouldn't have been L* 90 and used for buttons, and they shouldnt have severely limited chroma. In fact, I left over it and the dysfunction between VPs wondering why we didn't have it day 1, approving fixes repeatedly, and Android dysfunction that kept the conversation at "What? Didn't hear nothing from nobody in engineering! Anyways, lock screen clocks!"
> in 2020 built color system around the same science, enabling latest iterations of Material theming
No wonder everything Google builds, including Material, always has issues with contrast.
