Linear Light

Linear light is the encoding model in which signal values are directly proportional to actual scene luminance — twice the recorded value means twice the actual light from the scene. This is the natural physical relationship and the encoding used in scene-referred VFX work, particularly compositing and CG rendering. The OpenEXR format and ACEScg working space are both linear-light. In linear-light data, light addition, subtraction, multiplication and other compositing operations behave correctly — a 50% gray on top of a 50% gray gives twice the light value (a logical brightening), whereas in gamma-encoded data the math produces incorrect visual results. Linear light is, however, perceptually inefficient: human vision is highly nonlinear (we perceive much finer detail in shadows than in highlights of equal luminance ratio), so a linear encoding wastes a lot of code values in highlights and starves the shadows of precision. Display formats therefore apply a transfer function (gamma 2.4 for Rec.709, PQ for HDR) to encode for transmission efficiency, then linearize for math operations, then re-encode for display. The VFX pipeline typically does all heavy lifting in linear-light EXR files at 16-bit half-float precision, where math is well-behaved and there is no risk of banding from precision loss. The Filmic Blender and Academy Color Encoding System both rely on linear-light working spaces.