The simulator is 100% client-side. A colour is held as RGB state and transformed by pure functions in the page; an image is decoded by the browser, drawn to an offscreen <canvas>, read with getImageData, run through the colour-vision-deficiency matrices pixel-by-pixel, and re-drawn — all inside your tab. Nothing is transmitted: there is no upload endpoint, no server-side processing and no CDN involved in reading your image, and the built-in samples are hand-authored inline SVG so even they trigger no download. The tool works offline once cached, and closing the tab discards the image and the results.

This simulator applies the Machado, Oliveira & Fernandes (2009) model, which is derived from the physiology of cone response and is more accurate than naive channel-dropping tools, and it does so correctly — decoding sRGB to linear light before the matrix and re-encoding afterwards. Even so it is an AVERAGE-observer approximation for accessibility preview: individual colour vision differs from person to person, the brain adapts over a lifetime, and monitor calibration, gamut and brightness change every colour on screen. It cannot diagnose the type or severity of a deficiency — only a clinician using an Ishihara plate test or an anomaloscope can — and blue-yellow (tritan) simulation is the least reliable because screens reproduce blues and yellows least consistently.

Protanopia and deuteranopia are the red-green deficiencies: protanopia is a missing or non-functioning long-wavelength (red) cone and deuteranopia a missing medium-wavelength (green) cone, and together with their milder anomalous forms they account for the vast majority of colour blindness (around 8% of men of Northern-European descent). Tritanopia is a missing short-wavelength (blue) cone, a rare blue-yellow deficiency affecting far fewer people and both sexes roughly equally. Achromatopsia is the near-total absence of colour vision, where only luminance survives, so the image collapses to greyscale — this tool models it with a luminance-preserving greyscale. The severity slider blends each dichromat matrix back toward typical vision to represent the far more common ANOMALOUS (shifted-cone) versions rather than only the complete forms.

The WCAG 2.x contrast ratio uses relative luminance, and luminance changes very little under most colour-vision deficiencies, so the ratio shown here is essentially the same for a colour-blind viewer. That is exactly why contrast alone is not enough: two colours — a red and a green of similar lightness, say — can comfortably pass a contrast check and still be nearly indistinguishable to someone with red-green CVD, because what they lose is HUE discrimination, not brightness. The reliable fix is to never encode meaning in colour alone: add shapes, icons, labels, patterns or text, keep a strong luminance difference between items that must be separable, and use this simulator alongside a real contrast check rather than instead of one.