ACES Reference Gamut Compression 3D LUT

Hi we have not fully made the switch to OCIO V2 yet so cant implement this, We have show that they used this onset so struggling to match to the dailies colour for our QT deliverables for some shots where their is a lot of red lights.

I do have Jed Smiths nuke node but not sure how far off it is compared to the Official one as still a little off using this. Is there a 3D lut implementation I can try to see, it mentions about it in the docs. I know it wont be as accurate as the cpu/gpu implementation but need to try it.

Haven’t used it myself but @Derek has an implementation with a LUT. Contents | StudioX ACES Documentation Don’t know the exact details, accuracy and what the in/out space of the LUT should be but this may be what you’re looking for.

The cube in this is broke it appears not to be a cube at all but CLF wont load into OCIO File Transform and has lots of spurious text @Derek can you assist here?

In his config there is a LUT based option to implement in OCIOv1, how he does it can be seen from his v1 config file, and there is also a Nuke gizmo version but I can’t look at that since I don’t own Nuke. If that is non LUT based it’s the better option.

Got it now, it was corrupted on download for some reason

I see, glad you got it now. I had some time to look at it and test here. If the Nuke gizmo in folder \legacy-OCIOv1\software\Nuke doesn’t work the LUT way would be as follows if you’d want it as a node setup rather than implemented in OCIO config. The files are in the folder \legacy-OCIOv1\studio.

  1. Convert: ACEScg → ACES 2065-1
  2. Matrix / inverse direction
0.695452 0.140679 0.163869 
0.0447946 0.859671 0.0955343 
-0.00552588 0.00402521 1.0015
  1. File transform: aces_gm_lut_shaper_to_linear_v01.spi1d / inverse direction
  2. File transform: aces_RGC_65.cube
  3. Number 3 / forward direction
  4. Number 2 / forward direction
  5. Convert ACES 2065-1 → ACEScg

Hope that helps.
[edit] I now see that the matrix from the config is simply the conversion from AP1 to AP0 so you could remove step 1,2,6 & 7 as they cancel each other. The incoming data should be in AP1/Linear going into the shaper.

Thanks, it’s not as accurate as the Nuke gizmo but useful for adding to the viewer transform to get a better match.

1 Like

Howdy,

The LUT needs to have the the shaper as well. This is a 1D shaper function that transforms both negative and positive input value ranges into the 0.0→1.0 domain, essentially the ACEScct function mirrorred at the Y-axis.

However, even with the shaper the LUT is less accurate than the Nuke Gizmo, and is intended only for preview viewing. The best place to get the Nuke Gizmo from is here:

github /Joegenco/PixelManager

Uses OCIO v2.0 and includes ACES GamutCompression v1.3(From OCIO2.1) baked as a lut using ACEScct to carry the compression. Minor loss in LUT detail compared to OCIO 2.1 but quite close.

The problem is that ACEScct 0-1 does not cover sufficient range for the RGC. Out of gamut colours frequently have components with negative ACEScct values.