GFX-301e · Module 1

Color Space Precision

3 min read

When specifying colors for AI generation, the color space matters more than the value. RGB hex codes (#00ffff) are device-dependent — they look different on different screens without color management. HSL (hue 180, saturation 100%, lightness 50%) is more intuitive but suffers from perceptual non-uniformity — a 10-degree hue shift at H=60 (yellow) looks dramatically different from a 10-degree shift at H=240 (blue).

For AI color work, OKLCH is the most precise specification format. It describes color in terms of Lightness (0 to 1), Chroma (0 to 0.4+), and Hue (0 to 360 degrees) in a perceptually uniform space. A 10-degree hue shift produces the same perceptual change regardless of where on the hue wheel you start. Our brand cyan in OKLCH: L=0.905, C=0.155, H=194.77.

Why this matters for AI: when you set tolerance ranges for brand enforcement, perceptually uniform color spaces produce consistent enforcement. A +-5 degree hue tolerance in HSL allows wildly different perceptual shifts depending on the base hue. In OKLCH, +-5 degrees is +-5 degrees of perceived difference everywhere on the wheel. The enforcement is honest.

Do This

  • Specify colors in OKLCH for perceptually uniform tolerances
  • Use hex codes for final output targeting (web rendering) but OKLCH for specification and enforcement
  • Test color tolerances perceptually — does +-5 degrees look like the same amount of shift at every hue?

Avoid This

  • Rely solely on hex codes — they are device-dependent and perceptually non-uniform
  • Set uniform tolerances in HSL — a 5-degree shift at yellow is visually massive, at blue it is invisible
  • Ignore color space when evaluating AI output — the model generates in RGB, but you enforce in perception