Luttiehow to verify a lut is color-accurate (step by step)
a practical way to test whether a lut is accurate: run a colorchecker through it and measure the δe. the full workflow, and how to read the numbers.
a lut is a black box: it maps input colours to output colours, but the file itself tells you nothing about how it behaves. before you trust a lut on real footage — especially a technical or corrective one — it's worth verifying it. here's a repeatable way to do that without expensive gear.
why verify a lut at all?
there are two kinds of lut. a creative lut is supposed to shift colour — that's the look. a technical lut (a log-to-rec709 transform, a camera-matching lut, a corrective grade) is supposed to be accurate. the problem is that both arrive as the same kind of .cube file, and a technical lut that quietly introduces a green cast or crushes your shadows will ruin footage in ways you won't notice until it's too late.
verifying takes about a minute and turns "looks fine, i think" into a number.
the tool: a colorchecker chart
the standard reference is the 24-patch colorchecker — natural colours (skin, sky, foliage), saturated primaries, and a neutral grey ramp, each with a precisely defined value. you don't need the $80 physical chart to test a lut: grab the free digital colorchecker, which is generated from the official x-rite lab d50 reference values. (more on when digital is enough vs. a physical chart.)
step by step
- open the chart in your grading tool — or drop it straight into the luttie editor.
- apply the lut you want to test.
- check the neutral row first. the six grey patches along the bottom should stay perfectly neutral — equal red, green, and blue. any tint here means the lut has a colour cast, and a cast in the greys will show up everywhere.
- check the greyscale range. the white patch shouldn't be blown out; the black shouldn't be crushed. if detail is gone at either end, the lut is clipping.
- check skin and primaries. skin-tone patches should look natural, not orange or magenta. reds should read red, not orange; blues blue, not purple.
turn it into a number: measure δe
eyeballing catches gross errors, but "does the grey look neutral?" is subjective. to make it objective, measure the δe — the perceptual colour difference between each measured patch and its reference value. the modern formula is ciede2000.
the quickest way: export your graded chart and run it through the free color accuracy checker. it samples all 24 patches and reports the δe of each against the x-rite reference.
how to read the result:
- δe < 1 — imperceptible
- δe < 2 — excellent; the lut is accurate
- δe 2–5 — acceptable for most work
- δe > 5 — a visible shift you can see by eye
for a technical lut, aim for an average δe2000 under 2 with no single patch stranded far above 5. if the greys are clean but a couple of saturated primaries are high, that's usually fine — saturated colours are the hardest to hit and matter less than neutrals for most footage.
a note on creative looks
don't run this test on a creative lut and panic at a δe of 12. a film-emulation or teal-and-orange look is meant to move colour a long way from neutral — a high δe there is the point, not a bug. accuracy testing is for luts that claim to be faithful. for creative looks, judge intent, not distance from a reference.
do it once, trust it forever
verifying a lut is a one-time, one-minute check that pays for itself the first time it catches a cast before you've graded a whole project on top of it. grab the colorchecker, run it through the accuracy checker, and you'll know exactly what any lut is doing to your colour.