Hi,
Expanding on what I was talking about with respect to the Display-Rendering Transform (DRT) contrast during the 2021/01/27 meeting.
When creating CG assets, e.g. digi-doubles, generally, a VFX vendor will try to author them in a way that empowers him to reuse them across shows. With that objective in mind, the vendor will usually adopt a common working space and a DRT that will be used for the Lookdev of all the assets, independently of the show. It is not atypical for Lookdev or Lighting artists to only evaluate their CG assets or shots with the facility DRT, never use a client LUT and delegate any final look adjustments to the Compositing phase.
While excellent on paper, this approach is a double-edged sword that can bite hard, indeed, there is not really such a thing as a “neutral look” or “no Look”, and the DRT will inevitably affect the appearance modeling of CG assets.
Yesterday, @ChrisBrejon sent on the colour-science Slack a CG render of a digital head for evaluation of the Output Transforms which prompted a series of warnings, quoted here with minor editing:
- You are using the DRT to make artistic choices, thus you are biasing your render accounting for it, to give an example, and especially because here you have no reference for that particular lighting conditions, you are adjusting the specular response with respect to what you feel looks good not what is physically correct.
- Direct consequence of 1., if any issue occurs, and it will, because of the strong correlation between the DRT and the Lookdev, you won’t be able to properly disambiguate the source.
The takeaway here is that modeling skin appearance, is extremely hard and even with references, the DRT will have an effect on the artist judgement and in the hypothetical case where it does not, it could make artistic and technical assessments hard!
To continue with the skin example, I had long discussions in the past on the effect of the DRT on white skin and especially how it can be hard to tweak the specular response when the DRT contrast or highlights roll-off are so strong that it becomes hard to balance properly with the diffuse response.
As I was saying to Christophe, given a century of movies trying to make white actress look great, striving to reduce skin imperfections on both the hardware side, e.g. soft cine lenses, and the film processing side, it is not entirely unexpected! However, when the actress poses, without makeup, under the analytical lighting of an ICT-like rig equipped with razor-sharp camera lenses, the skin appearance changes quite dramatically. It becomes extremely difficult to reconcile the reference stills and motion-picture worlds, especially if the DRT layers a strong Look on-top of everything.
To fight that, I have seen a few talented artists wandering on extreme paths, disabling the DRT and the inverse EOTF entirely, letting the display decode an already linear image to help them peel-off the skin appearance with analytical precision. Before anybody screams with horror, I want to point-out that those artists are producing some of the best looking CG assets in the industry 
I’m hoping that this explains clearly a key reason as to why I would like a contrast reduction. I’m convinced that it would contribute improving CG assets appearance modeling, and also quite obviously help the colorist to do its work, but we have already talked about that point many times As I was also noting it, DRT contrast reduction is a real thing at some places, which given the advent of HDR displays is not entirely unexpected!
Cheers,
Thomas