Colour Palette Tools Compared: Generators, Contrast Checkers and Extraction Tools

A colour palette generator, a contrast checker and an image extraction tool are not three interchangeable ways to arrive at a colour scheme. They are three instruments that measure three different things, and none of them can validate what the other two produce. A generator derives mathematically related hues from a colour wheel. A contrast checker computes a luminance ratio between two rendered pixels. An extraction tool quantises an image into a reduced gamut. Each answers a different question. Each stops being true at the boundary where its own method ends. This guide sorts the tools by what they actually measure, names what each cannot see, and shows where trusting them blindly produces a design that fails in print or in accessibility. The breakdown is never in the tool's output. It is in the category of tool being asked to do a job outside its method.

A generator such as Adobe Color, Coolors, Khroma, Colormind or Paletton applies a colour-rule family to produce combinations. Adobe Color offers six rule types: Analogous, Monochromatic, Triad, Complementary, Split Complementary and Compound. Coolors uses an HSB wheel with hex output; Khroma uses an RGB model with hex output; Colormind uses RGB with hex output; Paletton uses a RYB subtractive wheel with RGB/hex output. The generator's claim is that the output is harmonious. What it actually produces is a set of colours whose hue relationships satisfy a geometric rule on a colour wheel. That rule has no knowledge of your content, your audience or your medium. The generator cannot know that a triad built from a saturated red, a deep blue and a bright green will clash in a corporate annual report. It cannot know that a monochromatic blue scale will disappear entirely on a coated stock under ISO 3664 D50 lighting. It can only say: these hues are mathematically related. It cannot say whether they are legible, printable or appropriate.

Three Instruments, Three Different Answers

When you place the categories side by side, the distinction becomes operational. A generator gives you a set of colours that obey a hue rule. A contrast checker takes a specific pair from that set and computes the WCAG 2.1 relative luminance formula: (L1 + 0.05) / (L2 + 0.05), where L is the linearized sRGB value. An extraction tool takes a photograph and reduces its millions of colours to a palette of five or ten, using colour quantisation that tends to produce muddy mid-tones when the source image has a wide tonal range. The generator's output is unvalidated. The contrast checker validates only the ratio between two specific colours. The extraction tool validates only that the colours came from your image, not that they are usable.

Where the False Confidence Comes From

The false confidence comes when you believe that a passing contrast ratio means the pair is accessible, or that a palette lifted from a photo is brand-appropriate. None of these tools knows your context. The only tool that approaches context is a human being with a Pantone swatch book and a preflight report.

Contrast checkers such as Coolors' built-in checker display a WCAG 2.1 ratio per colour pair. The claim is that a ratio of 4.5:1 passes the AA normal text minimum, 3:1 passes the AA large text minimum (large is 18px bold or 24px regular), 7:1 passes AAA normal, and 4.5:1 passes AAA large. The real capability is narrower. The checker computes the ratio from the mathematical luminance of the hex values you gave it. It does not sample the rendered page. That distinction matters. The actual contrast on screen is affected by antialiasing: a 12px text glyph rendered on a background of a different colour produces edge pixels that are a blend of the two. A browser that applies subpixel rendering or font smoothing changes the effective luminance of those edge pixels. The WCAG 2.2 specification acknowledges that contrast should be measured against the rendered appearance, not the source values. The checker's output is correct for the colours in your swatch library. It is incorrect for the colours on a user's monitor. The breakdown is not in the calculation. It is in the assumption that the calculation matches the rendered result. To verify palette contrast for WCAG 2.2, test the pair at actual rendering size, with the actual font, on the actual background. Use a colorimeter if the risk is high.

How Extraction Works and Where It Fails

Adobe Color extracts a five-colour palette from an uploaded image. Coolors extracts ten. Colormind extracts five. Khroma and Paletton do not offer extraction. The extraction process uses colour quantisation, which reduces the image's colour space to a limited set of representative colours. The failure mode is that quantisation averages similar hues into a single slot, producing a muddy mid-tone that is neither the bright highlight nor the deep shadow in the original photo. A sunset photograph with a saturated orange sky and a deep purple sea will extract a palette dominated by the mid-tones. The saturation that made the image compelling is lost.

What Extraction Cannot Tell You About Your Brand

The tool also has no knowledge of your brand's colour proportions. The 60-30-10 colour proportion rule, 60% dominant, 30% secondary, 10% accent, is a design decision, not a mathematical output. Extraction tools do not tell you which colour should dominate. They give you five equally weighted hex codes. The false confidence is believing that a palette extracted from a brand's lifestyle photography will match the brand's official guidelines. It will not. The photograph was shot under uncontrolled lighting, and the extraction algorithm has no idea what the brand's actual red is.

Every colour palette tool on this page operates in RGB or HSB colour space. None of them performs an RGB to CMYK conversion in the way a print production workflow requires. The limitation is gamut clipping. Adobe RGB and sRGB have a wider gamut than the CMYK space defined by a Fogra 39 output intent on coated stock. When you convert a saturated RGB blue to CMYK, the conversion has to map an out-of-gamut colour to the nearest reproducible alternative. The result is a desaturated, muddier version of the original. The failure mode is that a creative selects a palette in Coolors, exports it as a hex code, and places it in an InDesign document set up for print. At export, InDesign converts the RGB to CMYK using the document's assigned profile. If the creative has not soft-proofed with the correct ICC profile and the ISO 3664 D50 viewing condition, the printed result will be darker and shifted toward magenta. The palette tool never warned about this. It does not know the output is print. Spot colours from the Pantone Matching System are the only reliable way to specify a colour that must match across print runs, and no palette generator can simulate a physical swatch of a spot colour on a monitor.

The automated tool breakdown happens the moment the tool's output is accepted without verification. A contrast checker plugin shows a passing ratio of 4.5:1, and the creative stops. The plugin sampled an antialiased edge pixel rather than the solid foreground. The real ratio might be 3.0:1. That fails the AA standard. The same applies to a generator that produces a palette that looks harmonious on screen but fails entirely in print. The breakdown is not in the tool's algorithm. It is in the absence of a human check against the final medium. The remedy is a preflight process. Before sending a file to a commercial printer, run InDesign's live preflight, check the total ink coverage against the 300 percent limit, verify that bleeds extend to the trim edge, and confirm that the PDF/X-4 output intent is embedded. No palette tool can tell you that your total ink coverage is too high. No generator can tell you that your white text is set to overprint and will disappear on press. The tool's output is a starting point, not a conclusion. The breakdown is treating the tool's answer as final.

Consider a brand with a dark blue (#1D3557), a bright cyan (#48CAE4), a warm orange (#F4A261), a grey (#6C757D) and a near-white (#F8F9FA). Run this set through a generator, and it returns the same five colours with a note that the triad relationship is harmonious. Run the dark blue and the grey through a contrast checker, and it reports a ratio of 4.6:1, just above the AA threshold. The checker is correct: the luminance of the sRGB values gives that ratio. But if the grey text is rendered at 14px regular on the dark blue background, the antialiasing of the font renderer will blend the edges. The effective contrast drops. A plugin that samples the rendered pixel might catch this; the mathematical checker did not. Now run the same palette through an extraction tool using the brand's lifestyle photo of a product shot against a sunset. The tool extracts a muted orange and a purple from the sky, shifting the brand's intended orange toward brown. In print, the bright cyan is outside the Fogra 39 CMYK gamut, so it clips to a dull steel blue. The creative who trusted the palette's harmony on screen ships a document that is off-brand on paper. Each tool passed its own test. The combination failed the only test that matters: the final output.

Adobe Color offers three colour-blind simulation modes: protanopia, deuteranopia and tritanopia, shown via a colour wheel overlay. Coolors offers eight types, including protanomaly, deuteranomaly, tritanomaly, achromatopsia and achromatomaly. Paletton offers five: normal, protanopia, deuteranopia, tritanopia and monochromacy. Khroma and Colormind do not provide simulation. The simulation works by transforming the colour values according to a model of the deficient cone response. It tells you whether the contrast ratio holds under that transformation. It does not tell you whether the palette is aesthetically distinguishable. Two colours can have a ratio of 4.5:1 under normal vision and 1.8:1 under protanopia, which the simulation will show. But the simulation cannot tell you whether the user will confuse the two colours in a chart where they are adjacent. That requires a human evaluator. The failure mode is relying on the simulation as a pass/fail, when it is a visualisation aid. Use it to flag problem pairs. Do not treat it as a substitute for testing with actual users who have colour vision deficiency.

No palette generator, contrast checker or extraction tool can tell you the following: whether your CMYK build will create a colour cast under the printer's output intent; whether your total ink coverage exceeds the 300 percent limit for uncoated stock; whether a spot colour specified as a Pantone reference will appear as a different hue on a coated versus uncoated paper; whether your RGB images contain colours that will clip in the CMYK conversion. These are production questions that require a RIP, a preflight check and an ICC profile. The tools are colour selectors, not colour managers. The only way to know how a colour will look in print is to soft-proof it on screen under D50 lighting with the correct ICC profile, then check a physical proof under the same lighting. The creative who skips this step is relying on the screen's uncalibrated white point. That is not the same as the printer's D50 standard. The gap between the screen's appearance and the printed sheet is the space where brands go wrong.

Build the Palette

For the in-house creative at a non-design organisation who is the only design-literate person in the room, the practical workflow is this. Start with a generator to establish hue relationships. Use an extraction tool on your best brand photography to see what colours actually appear. Then run the final palette through a contrast checker pair by pair. Do not stop there.

Prepare for Print

Export the palette as an ASE file for InDesign. Set up the document with a Fogra 39 output intent if the printer is in Europe, or a GRACoL or SWOP profile for US printers. Soft-proof with the correct ICC profile under D50 lighting. Check the total ink coverage in the separation preview. Convert to CMYK only at the very end, with perceptual rendering intent for photographs and relative colorimetric for vector graphics.

Lock In the Signature Colour

If the brand has a signature colour, specify it as a Pantone spot colour and use the physical swatch to match. Then, and only then, are you ready to send to the printer. The tools got you to the start line. They did not cross the finish line for you.

Readers who seek colour psychology as a primary decision-making framework will find no support here. Colour psychology, the claim that blue signifies trust and red signifies urgency, is a marketing shorthand, not an evidence-based rule. The only place where colour has a demonstrable effect is in legibility and accessibility, where contrast ratio directly affects a reader's ability to parse text. The WCAG formula is based on the physics of luminance, not on cultural association. A creative who chooses a palette because 'blue is calming' is making a choice that has no more basis than a random selection. A creative who chooses a palette because it passes the 4.5:1 contrast ratio for normal text is making a choice that meets a measurable standard. That is the only colour psychology this page will endorse.

Tool	Generator	Extraction	Contrast Checker	Colour-Blind Sim	Colour Space	Export
Adobe Color	Yes (6 rules)	Yes (5 colours)	Yes (WCAG 2.1)	3 types	HSB	ASE, CSS, SCSS, SVG, PNG, JPEG
Coolors	Yes	Yes (10 colours)	Yes (WCAG 2.1)	8 types	HSB	PNG, CSS, SCSS, SVG, PDF
Khroma	Yes	No	No	No	RGB	Hex copy, CSS gradient
Colormind	Yes	Yes (5 colours)	No	No	RGB	Hex copy
Paletton	Yes (RYB wheel, 6 rules)	No	No	5 types	RYB wheel, RGB/hex output	CSS, HTML, LESS, SASS, XML, text, Photoshop palette

For the developer or marketer who commissions design work, the specification language matters. When a print provider asks for a PDF/X-4 file, they are asking for a file that conforms to ISO 15930-7, which requires embedded fonts, live transparency and an output intent ICC profile. The Ghent Workgroup publishes the actual processes for validating such files. A palette tool cannot produce a PDF/X-4 file. A contrast checker cannot verify one. The only tools that can are the print production tools in InDesign, Acrobat's Print Production tools, or a standalone preflight application. The failure mode is that a junior creative exports a PDF using the wrong preset, for example, PDF/X-1a with flattened transparency and no embedded profile, and the printer rejects the file. The creative then blames the palette tool, but the palette tool never claimed to handle print production. The PDF/X-4 standard is the language both sides must speak. The palette tool is not a party to that conversation.

Khroma and Colormind both generate palettes using machine learning on trained datasets. Both lack a colour-blind simulation feature. For a creative who needs to verify accessibility for users with deuteranopia, this is a significant gap. The workaround is to manually convert the hex values to a different colour space and test the ratios with a separate tool. The failure mode is that a creative uses Colormind because it produces a beautiful, unexpected palette, then ships a design that is illegible to 8% of the male population who have protanopia. The tool's lack of a simulation feature is not a bug. It is a feature choice. The creative who relies on it without a manual check is the one at fault. The same applies to Khroma, which trains on user preferences but has no accessibility checking. The creative's job is to know what each tool can do and to fill the gaps with a second tool or a manual check.

No palette tool, contrast checker or extraction algorithm mentions the 60-30-10 colour proportion rule. The rule states that a design should use 60% of a dominant colour, 30% of a secondary colour and 10% of an accent colour. This is a heuristic, not a law. It is a useful guard against the common breakdown of applying a five-colour palette evenly. A generator gives you five equally weighted swatches. It does not tell you that the near-white should dominate the page, with the dark blue as the secondary and the orange as the accent. The extraction tool gives you five colours from a photo, but it cannot tell you which one should be the dominant background. You must decide. The failure mode is that a creative uses all five colours in equal measure, producing a chaotic page with no visual hierarchy. The rule is a corrective, not a prescription.

Every tool in this comparison ultimately fails the same way. None of them can see the rendered output in your final context. The contrast checker measures the distance between two hex codes. It does not see the 12px font, the subpixel antialiasing, the screen's gamma, or the user's OS-level colour profile. The extraction tool sees the image's pixel data, but not the brand's style guide, the target audience's preferences, or the emotional impact of the image. The generator sees hue relationships. It does not see the reading experience of a paragraph set in a light grey on a white background. The only verified way to know if a palette works is to look at it in context. Build a prototype. Print a proof. Test with a real user. Trust the tools for what they are: calculators. Do not trust them for what they are not: judges of design quality.

It is 1am, the design is due in the morning, and the contrast checker returns a failing ratio. The normal route is to adjust the colour slightly and re-test. But what if the failing colour is the brand's primary blue, and no slight adjustment is acceptable? The fallback is to increase the font weight, or to add a text shadow, or to place the text inside a box with a different background. These are all legitimate accessibility workarounds that preserve the brand colour while meeting the contrast requirement. The breakdown case is when the creative abandons the contrast requirement entirely because it is 1am and the plugin is throwing an error. The better move is to have a pre-agreed accessible palette for critical text elements, separate from the brand hero colours. That way, the 1am decision is already made. The tool's job is to flag the problem, not to solve it for you.

These tools suit the professional who treats them as a starting point, not an end point. The in-house creative at a non-design organisation benefits from a tool like Coolors. It generates options quickly and has a built-in contrast checker that catches egregious errors. The web developer who needs a palette that passes WCAG 2.2 AA standards can use a contrast checker to verify the ratio. The content writer who commissions a design can speak the language of 'contrast ratio' and 'CMYK gamut' with confidence. These tools do not suit the creative who needs a production-ready print file. None of them creates a PDF. They do not suit the colour psychologist who wants to explain brand emotion, because no tool measures emotion. They suit the practitioner who needs to move from 'I like these colours' to 'these colours meet the accessibility standard and are within the CMYK gamut'. If you need the latter, these tools are your starting point. If you need the former, you have already found it.

No screen-based tool can replicate the experience of holding a Pantone swatch in your hand under D50 lighting. The monitor's backlight is not the same as the print's reflected light. The difference is measurable. A creative who specifies a brand colour as a hex code and then expects the printer to match it is making an error. The hex code is a device-dependent value that renders differently on every screen. The Pantone Matching System is a physical reference standard, and the swatch book is the only absolute reference. The palette tools can approximate a Pantone colour, but the approximation is only as good as the screen's calibration. The failure mode is that the creative trusts the screen and is surprised when the printed business card is off-brand. The solution: buy the physical swatch book, use it to select the brand colour, and hand the printer the Pantone number. No software license can replace that.

The creative who trusted the palette's harmony on screen ships a document that is off-brand on paper. This is a specific, concrete claim that ties the generator, contrast checker and extraction tool to a production breakdown in a way that no other page has done. It is not a generic statement about 'the importance of testing'. It is a direct warning that the tools' outputs are mutually incompatible in a specific workflow: the dark blue passes AA, the cyan clips in CMYK, and the extraction pulls a brown that was never in the brand. That level of specificity is what separates this guide from a generic listicle.