This is an ACES 2.0 question for those in the know about colour science. I saw you replied to another thread recently @doug-walker so I would be interested in your expert thoughts on this.
I have a project that is ACES 2.0 but they are insisting that an ACES 1.3 RGC is used within the ACES 2.0 specified colour pipeline. I thought one of the main reasons behind ACES 2.0 was to avoid having to do any gamut compression, and that the transforms had been adjusted to account for that? My concern is that we would effectively be doing what could be considered a double compress.
Does anyone have any “best practice” information on this? There are several threads on the ACES forum but nothing definitive and some threads contradict others.
I’m surprised that ACES 2.0 didn’t include a wider gamut than AP1 with a fair few cameras capturing beyond this. Totally fine as I am yet to see a compromised end result and P3 is smaller than AP1.
RGC = Reference Gamut Compression for those not in the know (I got asked). It was introduced in ACES 1.3 to deal with highly saturated, out of gamut, colours and bring them into the AP1 colourspace in an aesthetically acceptable way, with some roll-off (i.e. non-linear).
Hey Adam,
It’s a good question, and others may have more practical experience, but I will share my thoughts on this. As you know, ACES 2 is much more robust that ACES 1. With ACES 1, it was not unusual to encounter saturated colors that would just totally break. I don’t really experience that with ACES 2, but as with any DRT, there are occasional artifacts.
I would say that RGC is not essential with ACES 2, but it can be helpful on certain shots. The gamut compression in ACES 2 uses the concept of a “reach” gamut which is explained at this link on the ACES site. The reach gamut is AP1. Colors outside the reach gamut are essentially clamped along a hue-preserving line towards neutral, based on the cusp at that hue.
You are certainly correct that most camera color space gamuts are wider than AP1. The ACES RGC was designed to bring the output of digital cinema cameras into AP1. So it is the case that very saturated colors in a camera gamut that would otherwise clip to the reach gamut would have additional detail preserved by using RGC.
Whether that is good or bad will definitely depend on the shot. I’ve seen shots where the RGC was helpful with ACES 2 and other shots where I felt that it caused unhelpful desaturation. The desaturation is especially noticeable when working on an HDR display. Your concern about double compression is definitely justified! But the client is also correct that it will preserve detail in very saturated colors on certain shots and they may want to err on the side of avoiding clipping than preserving saturation. Of course, most shots won’t have sufficiently saturated colors be able to tell the difference either way.
In summary, the RGC is still occasionally useful, even with ACES 2. It’s probably something I would apply as needed rather than on every shot, but it does not surprise me that some people may want to leave it on. As long as they are aware of the trade-off that you raised, I think it’s ok.
(For those reading this thread, it may be helpful to note that in the default Flame OCIO config, ACES RGC is available as a Look, which means it may be applied in Flame’s Look Batch node or timeline FX, or as part of a custom Colour Transform.)
Doug
Thanks @doug-walker
Your answer does clarify a lot of things to me. I was under the impression that everything would map into AP1 using the DRT but if there are occasions when that isn’t the case then I wasn’t aware, so I now feel better informed.
My thoughts would be that if you were going to apply RGC that you would do so in the grade in DI if absolutely needed, not at the very start of the process. It was suggesated that any resizing/repositioning could create more out of gamut colours (which makes sense when you take filtering into account) so they want it at the very start. I feel like the RGC should correct that at any stage but I didn’t want to get into a technical tit for tat as that generally turns into a battle of ego rather than being constructive. It was a case of best practive vs something that still works argument so not worth having unless there would be a serious compromise. Looking at the post RGC image, the desaturation wasn’t to the point of being horrible or problematic on this particular production.
We have adhered to the colour pipeline request as it is a bit of a grey area but I do feel justified raising my concerns so thank you. I appreciate your taking the time to respond so quicklly.
Yes, I agree it’s not ideal to bake it in at the start of the process. In fact the RGC User Guide on the ACES site explicitly says: “Frame pulls for VFX should NOT have the Reference Gamut Compression baked in.” It does have the benefit of moving everything inside AP1, which could be helpful if ACEScg or ACEScct is being used as a working space, but yeah, it’s not clear whether resizing would be better before or after.
As you say, it’s something that has pros and cons and could be done either way the customer prefers.
Once again, thanks @doug-walker
Sounds like we are on the same page as to what we think the correct workflow should be. In this particular circumstance the client feels different so will stick to their suggested workflow, even if I don’t agree with it.