Color arithmetic must be done in linear light, not in gamma-encoded values

Gamma-encoded values (sRGB, DCI-P3, etc.) are compressed for storage, not for math. When you mix or interpolate two gamma-encoded colors directly, the result is incorrect — midtones appear darker than physically accurate. Fix: linearize (apply the inverse TRC) before arithmetic, then re-encode afterward. In GLSL shaders, texture samples tagged as sRGB are often auto-linearized by the hardware; if not, pow(color, vec3(2.2)) is a common approximation. This is the root cause of ‘muddy gradients’ in naive color interpolation.

Examples

Linear gradient red→green: the middle step has equal physical red and green, perceived as yellow. sRGB-encoded midpoint computed naively: physically darker, perceived as olive/brown.

Assessment

A shader lerps between two sRGB colors and the midpoint looks too dark. Describe exactly what went wrong and how to fix it.