Hi im wondering if this is the correct workflow for roundtripping between Davinci (YRGB) and Cinema 4D for 3D camera tracking and VFX work in 2024/2025:
In Davinci
→ Footage (RED Wide Gamut / Log3G10)
→ Color Page: ACES Transform node: ACES 1.3 - Input: DWG/Log3G10 - Out: ACEScg
→ Delivery Page: Export EXR (RGB half DWAB), Data LVL Full, Color Space Tag: ACES AP1, Gamma Tag: Linear
→ In Cinema 4D: Project settings/Color management: Render Space ACEScg (“…Colors managed by OCIO”, “Calculations happen in ACEScg…”, “Colors interpreted in ACEScg…”)
→ Export with AOVs from Cinema 4D using Redshift: Multi Layered EXR (OpenEXR) 32bit
→ Imported into Davinci Resolve Fusion Tab: Nodes: Loader to ACES Transform (1.3 / In: ACEScg / Out: Working Color Space) to Compositing nodes to ODT node if done otherwise ODT after grading in Color Tab.
Does that seem correct? Also should i better stay within ACEScg while compositing in Fusion or is DWG Linear the better way since its native to Davinci?
I personally don’t see any issues with the pipeline.
DWG Linear is not any more ‘native’ than the other color spaces they have available. From a practical point of view I would say use the camera linear color space while compositing if a camera is involved. If it’s only CG renders in ACEScg just keep that. There is no benefit of using DWG. The risk you run into with using other working spaces is creating negative values. Problematic if you have to manipulate them. Since DWG is larger than ACEScg it doesn’t matter for that context but it does make your ‘raw rgb colors’ you generate in that working space much more sensitive since the primaries of DWG are further out in imaginary color land.
If you use RCM DWG/Intermediate for finishing you don’t really get to choose the working space in Fusion, it will be the linear version of the timeline color space, but you could convert that to the camera linear space or ACEScg within Fusion. The same is true for ACES based management.
I would personally recommend completely manually rolling Resolve’s management however as that gives complete control over what happens when without the restrictive implied workflow you don’t have any agency over. Also keep in mind that when using CST nodes to convert scene to scene you disable the tone mapping, this tends to be overlooked.
Hey thank you so much for your detailed and helpful answer Shebanjah!
Id love to hear more about the last thing you said. I am manually color managing each scene/clip in Davinci per CSTs and i guess im guilty of overlooking the fact that tone mapping is disabled. Can you elaborate on that? I feel like i dont really understand what you mean by that.
Sure! If you enable tone mapping at any stage a conversion needs to happen to a space that is not a display, the transformation is not scene referred anymore. So converting any combination of log/linear spaces would still apply tonemapping to the luminance which is defined by the maximum input and maximum output nits of the CST. There are default values in place depending on the chosen in or out gamma that the user can override by checking custom. But it should not be part of a conversion unless it is either the final stage of converting to the target display for the delivery (just like ACES’ ODT output transforms) or if you intend to invert a display transform to scene like ACES ODT inverses can do as well. Such function is primarily used to preserve already color rendered/normalized images appearance for the same display deliverable. Think logos, stock footage, graded material that is already Rec.709 → inverse DRT → working color space → forward DRT → is the same Rec.709 image again.
The way RCM handles this is that the user can define if it wants an inputDRT which is responsible for mapping input to timeline space. Using the same as the outputDRT creates this mechanism which would cause Rec.709 tagged video to appear the same if that’s also the output. This is one of the reasons I’m against the use project wide management in Resolve because you don’t get to dictate which media gets such treatment and which don’t. You do not want this input to timeline tone mapping for camera data or CG renders. There are many other quirks in RCM but that’s a bit outside of the discussion I think…
