Nuke Linear Color question

Here’s a fun one:

Any time you are working on linear footage, you’re looking through a viewing lut, so it stands to reason that something that looks correct under one viewing lut will look incorrect under another.

The idea behind ACES (or one of them) is that everyone can look at the same image and see the same thing.

But i’ve been seeing a lot of jobs that are being done using Nuke’s default “linear” viewer lately.

You can make this viewer in flame by picking the lut preset “Scene Linear to Video via sRGB”

The problem I keep running into, is the flavor of Scene Linear that this lut displays correctly does not correlate to anything available to us as flame ops on the input transform menu, which means there is no “correct” way to convert the image from Nuke-Scene-Linear-sRGB to, say, ACES.

And I feel like I’m missing something huge somewhere, but I cannot get the imagery to look the same under the ACES (or the Arri) viewing luts.

As I noted before, I can VIEW this just fine, but converting it into any other format makes it look different (broadly speaking, ACES and Arri viewing luts render the image darker), which makes me insecure about doing that at all.

So does anyone know what the right way to crack this would be? I just want my imagery to look correct in ACES so I can then convert it to whatever the heck I want.

Having access to mathematical expressions in 1d lookups would at least be helpful in recreating that particular curve (or reversing it) in a meaningful manner within the construct of the color management node. Is there anywhere one can do per pixel math on an image… and by do math I mean mathematical expressions akin to what’s available in the channel editor? Seems like someone would have created a matchbox node somewhere, then you could preprocess…

The srgb lookup is x<.04045 ? x/12.92 : pow((x+.055)/1.055, 2.4) for the record. First bit’s an if/than followed up with an else.


1 Like

If my memory serves me well you should use standart gamma 2.2 instead sRGB for interpretation in Flame (it`s not correctly named in Nuke).

Other part of a problem is ACES workflow itself. For viewing (and outputing in some cases) process ACES uses not straight conversion from ACEScg (or ACEScc) to whatever you need but also an RRT (reference rendering transform) - think of it like some kind of embeded look. It can definшtely affect viewing experience