We are experiencing some confusion going from Nuke to resolve, and comparing the resolve output with what we see in Nuke. We are experimenting with ACES for the first time, and are by no means very experienced.
We render our CG to the ACEScg space
renders are ingested in Nuke as ACEScg
Comp work is converted to ACES2065-1 and written to 16bit aces compliant EXR
In Resolve Project settings, color science is set to ACEScc
IDT is set to No Input Transform (We don’t see an ACES2065-1 IDT), ODT is set to Rec.709
Prores mov is exported and viewed RAW in RV (no image transform and no display transform)
Compared alongside the Nuke comp, The colors are off (its a render of a macbeth color chart alongside some shaded objects). The whitepoint looks fine, but the pink, red, blue colors are notably desaturated.
Question: Am I looking at a disparity in Primaries?
Question: What is the correct workflow from Nuke to Resolve and Inside Resolve itself, to faithfully retain the comp color decisions for Rec.709, DCI P3, and Rec.2020 (our various deliverables)
Observation: If we use ODT of rec.709(D60 sim.) and write that out, the colors match a lot better, but the whitepoint is now warmer- which is not what we would want either, and we’re not sure of how to deliver for DCI P3 since it doesn’t have a ‘D60 sim.’ ODT
Can you post sample EXRs before and after Nuke? What you describe should work
No IDT is needed for ACES2065-1. An IDT is a transform to ACES2065-1, so if the image is already in that colour space then none is needed.
The DCI ODTs are ‘D60 sim’. This is just not made clear by the ACES 1.0.3 naming convention. Name changes to make this clearer are part of the ACES 1.1 update.
Nuke and Resolve match for me. I don’t have RV, so can’t test that, but when I bring a ProRes render from Resolve back into Nuke it matches the original EXR + the Rec.709 ODT.
Have you tried bringing the ProRes back into Nuke? Set the Read node colorspace to “Output - Rec.709” so that with “Rec.709 (ACES)” as the VLUT it’s essentially a passthrough.
We’ve investigated the issue further and have no idea where we went wrong the first time round.
With trying it again on 2 Different machines, everything seems just fine.
Thank you for your help though. We look forward to seeing the ACES 1.1 update in OCIO.