I was referring to the side by side images of @Jean-Charles_Fouche. The main question should be about colour accuracy and keeping the workflow colour managed. My remark is that the display-referred appearance of colour at the intermediary steps should not matter and a side by side comparison irrelevant as long as the data is intact and the profiles are correctly done.
Regarding the first point, that is not the reason that we don’t use ACES as the input profile. The input profile should be the camera’s profile, either stock profile or manually profiled profile. Once specified, RT will know what to do to convert the data to the working profile according to RT’s extended Adobe coefficients info.
Selecting the output profile is the most straightforward step, as you should be able to set it to anything you wish. In particular, since ACES AP0 is so large, you don’t need to worry about gamut clipping or compression.
Notice how I skipped working profile just now; because this is where it gets complicated. It isn’t as simple as making it the same as the output profile and your job is done. That is okay only if there are no other subsequent operations. I don’t know if RT’s neutral preset has nothing in the pipeline after the working profile is appended. It would also depend on what you plan to do at the ingest stage prior to exporting to Resolve.
You may ask, What is the fuss if my working profile is sufficiently large and the output profile is the same so no final conversion takes place? There are things to discuss but they are out of my depth. The thing to note is that if the working space is too large or too small or the profile is malformed, you could have problems with artifacts or discontinuities. Moreover, not all operations are compatible with all spaces.
If you insist on using ACES, note that ACES is mainly a storage format. That is a monkey wrench for you. @XavAL mentioned @Elle’s profiles. You would have to double check but if memory serves her ACES profiles should be AP1 not AP0. This is for the explicit reason that I described above: AP0 is for transport and too large of a colour space is no good. There are other more specific ACES specifications but that is over my head again.
The takeaway is that at this point we are getting in the weeds. The topics discussed might not matter to us regular folk. The reason behind all reasons is that ICC profiles, CLUTs and matrices are limited and don’t sufficiently describe the psycho-physical properties of colour or how colour is designed to be stored and processed digitally. It is a broad and complex subject that is still being researched. E.g., characterizing a camera’s actual spectral sensitivities and implementing them into the workflow.
Certainly, we can do our due diligence, learn from one another and control our workflow as much as possible but then we need the resources, people, expertise and close collaboration to accomplish that, especially in the cases where the project is expansive like those in Hollywood.