How this works and what it does not do
The limits below are not disclaimers bolted on at the end. They are the shape of the product.
Last updated 10 September 2026
01What the service does
You give it a set of colours and a short description of where they are going to be used. It pairs them — every colour against every other colour — and works each pair through several situations that commonly change how two colours relate to one another: a greyscale conversion, a projected image, a screen in bright light, a dark theme, and one common type of colour vision.
For each pair it reports which of those situations the two become the same colour in, what information you told it was relying on that pair alone, the measured values, and what you can do about it without changing either colour.
The calculations are ordinary arithmetic and they run in your browser. Relative luminance and contrast ratio are defined in WCAG 2.1. The colour difference is CIEDE2000. The colour-vision transform is a published matrix from the peer-reviewed literature. None of these involves a generative model, and no generative model produces any part of a result: an approximation invented by a model would look convincing and would support confident wrong conclusions, which is the opposite of what this is for.
02It makes no judgement about whether anything meets a standard
The service gives the measured contrast of a pair and names the threshold a standard sets. That is a numerical comparison, not a conclusion.
Whether something meets a standard depends on size, weight, use and whether the same information is carried in any other way — none of which the service can see.
The service therefore issues no compliance opinion, no score, no grade and no pass mark, and its output must not be used as the basis of an accessibility audit, assessment or statement. If you need that, you need someone who can look at the whole piece of work.
If you ask it a question about compliance, auditing, certification, a legal requirement or an accessibility statement, it will not answer the question in a hedged way. It will decline the whole request and say so.
03The scene renderings are approximations, not anybody's eyes
The colour-vision rendering is an approximation produced by a standard transform. It is not what any particular person sees.
Colour vision varies from person to person, there is more than one type, and it varies in degree. The service models one common type at one severity because that is the case most likely to catch a pair of chart colours. It is a reason to look again, not a report on anyone.
The service does not assess, infer or diagnose anyone’s colour vision or eyesight, and it provides nothing relating to eye health. Questions about your own vision belong with an optometrist. If you ask it something about your own eyes, it will decline the whole request and tell you that.
04It cannot promise what anything will actually look like
The service cannot guarantee how anything will look when it is printed, projected or seen in daylight. Paper, ink, toner, the machine and its own internal conversion, the projector, the screen surface, the lighting in the room and the ambient light outside all change the result, and none of them are visible to this.
Every situation in the product is shown with a list of what it cannot know, printed next to the result rather than hidden in a footnote. Treat the output as a way of finding the pairs worth looking at, then look at them.
05It generates no palette and gives no replacement colours
No colour palette is generated here and no replacement colour value is given. The service may describe a direction — darkening one of a pair to open up the lightness difference, for instance — but it does not design anything for you, because it knows nothing about the rest of the work. It makes no aesthetic judgement: nothing here calls a colour good, bad, dated, professional or on-brand.
The course the service recommends first, in almost every case, is not to rely on colour alone. Once a pattern, a direct label or a line style also carries the distinction, it no longer matters whether the two colours merge — and no conversion can take that away.
06Accuracy, and what you are responsible for
The measurements are exact for the values you supply. The judgement built on top of them is not: calling two colours “merged” uses a threshold, a threshold is a choice, and so the measured difference is printed in every cell for you to apply your own.
You remain responsible for the work you produce. The output is information to inform a decision, not the decision, and not professional advice of any kind.
07High-risk uses are not permitted
The service must not be used as an input to medical, legal, financial, safety-critical, employment, credit, housing or judicial decisions. See the acceptable use policy.
08What happens to what you type in
The colour values and the short description you type are not used to train any model. There is no image upload on this service, so it never receives your file, your chart or your document. If you have an account, sets of colours you save are stored so you can return to them, and you can delete them at any time. The privacy notice sets out the detail.
If any of this is unclear, write to support@picassod.shop and ask.