I simulated the colour vision deficiencies here using SVG filters replicating the Brettel-Viénot-Mollon algorithms. It isn’t completely perfect, but I’m pretty happy with the approach, as it avoids creating, storing, and loading multiple variants of the exact same image.