Free tool · nothing is uploaded

Are these two LUTs actually different?

A number instead of a guess. Both files sampled on a shared grid, with the largest disagreement reported in the units your footage is actually quantised to.

First LUT

Drop a .cube

  • .cube

Read on your own device — nothing is uploaded.

drop, click, or paste

Second LUT

Drop a .cube

  • .cube

Read on your own device — nothing is uploaded.

drop, click, or paste

What the number is measuring

Both LUTs are asked the same set of questions — given this colour, what comes out — across a grid covering the whole cube. The reported figure is the largest disagreement found on any single channel at any of those points, scaled to the 0–255 range an 8-bit file is quantised to.

A worst case rather than an average, deliberately. An average would quietly pass a LUT that matches perfectly across skin tones and diverges hard in saturated greens — exactly the failure worth catching before you grade forty minutes of footage with the wrong file.

DifferenceWhat it means
< 1Below one 8-bit step. These cannot produce a different pixel — it is one LUT under two names.
1 – 3Visible only on a wide smooth gradient. On footage, the same look.
3 – 12Real but subtle — a look and the same look dialled back, or two grid sizes of one transform.
> 12Different looks. Obvious on any image with colour in it.

The question this was built for

LUT packs are sold by the count — forty cinematic looks, sixty film emulations — and the count is not always honest. A pack assembled by exporting one grade at varying opacity produces forty genuine files that measure as a smooth ladder of small differences, and no thumbnail grid will show you that.

It answers quieter questions too. Did the LUT that came back from the colourist actually change, or did it round-trip untouched? Is the 17³ export close enough to the 65³ to send to the on-set monitor? Both are a number, and both are otherwise a guess.

Next steps

Frequently asked questions

What does the number mean?

It is the largest disagreement on any single channel, at any colour, expressed in 8-bit steps out of 255. A value of 2 means that somewhere in the colour cube one LUT outputs a red, green or blue two steps away from the other — and nowhere does it differ by more. It is a worst case rather than an average, which is the useful shape: an average hides a LUT that matches everywhere except in the deep blues.

How can it compare a 17³ against a 65³?

By sampling both on a shared grid rather than lining entries up against each other. Comparing entry by entry would be meaningless, since the files have different numbers of entries in different places. This asks both LUTs the same set of questions and compares their answers — which is also why a downsized LUT measures as nearly identical to its source rather than as a different file.

Below what number are two LUTs the same?

Under 1.0 they cannot produce a different pixel in an 8-bit file, because the difference is smaller than one quantisation step. Under 3 you will only see it on a wide smooth gradient. Above about 12 the difference is obvious on any image with colour in it.

I bought a pack of 40 LUTs. How do I check they are not all the same?

Compare them in pairs, starting with any two that look similar in the thumbnails. A pack built honestly shows double-digit differences between its looks; one built by exporting a single grade at varying opacity shows a smooth ladder of small numbers, and the strength tool tells you the same story from the other direction.

Why not just look at the two previews?

You should do both, and the tool shows both. But a preview is one image, and two LUTs can be indistinguishable on a portrait while diverging badly in saturated greens or crushed blacks. The number covers the whole colour cube, including the corners your sample frame never visits.

Are my LUTs uploaded?

No. Both files are read with the browser's File API and compared on your own device. Nothing is transmitted and there is no account.

A LUT tells you nothing about how it was built.

That is the trade you accept when you bake one: thirty-five thousand answers and no record of the question. The Studio works the other way round — give it a photograph and its recipe and it shows the edit as a sequence of decisions, each one rendered on its own.

Open the StudioFree, no account, nothing uploaded.

Related