Andrew Kolakowski wrote:Atm. it's all mess and I see not much hope for a change as it's nothing which can bring money, so no one is interested
I agree with you, it is definitely a mess. The main thing I realize is for broadcast work or streaming delivery, the tags don't matter (no streamer asks for any specific NCLC tags in specs, for instance). Just grade through Ultrastudio into 3D LUT based rec709 2.4 gamma monitor (like my Flanders DM240) and that's it!
But with ICC / ColorSync systems like NCLC tags and color managed apps like Quicktime, there seem to be some really odd behaviors. Firstly, my understanding of the system, which I tested with multiple tags, using matching RCM workflow, is that the end result shows the same (except for any out of gamut color or out of luminance values, in the case of HDR ) in the color managed app. So, for example, if I use RCM and select rec 709 2.4 and render, it will look the same as a file set to rec2020 2.4, in Quicktime (except any out of gamut colors, which will be dealt with according to how you have Resolve set up). But the understanding I have, is that Quicktime sees the tag and changes the viewing conditions to match. If you output from a RCM workflow, it will look identical (again except for out of gamut / out of luminance in HDR). This behavior seems great, as it smartly transforms your monitor for you. Where as, if I bring the same files into Resolve and turn off color management, the rec709 one will look different than the rec2020, which is the true nature of the file, and would require manually calibrating the monitor to each standard.
However, with this ICC / colorsync based system, there doesn't seem to be a known monitor space / gamma, starting point / monitor profile? So you have no way of knowing if the final image is absolutely accurate, to a known calibrated source. See the problem with this? So when I tag my image with rec709 2.4 (1-2-1) it may have the "correct" tag, but the resulting image looks relatively dark compared to my Ultrastudio output into my 3D LUT calibrated Flanders DM240 set to Rec709 2.4. So this can't be the correct viewing environment. (Let's also set aside the concept that the NCLC is defining the EOTF and not the OETF, which doesn't seem to be the case, at least for rec709, according to the ProRes document I linked in a previous post).
For example, I ran the HDMI output of my Mac Studio into the DVI input of my recently 3D LUT calibrated DM240 (set to Rec709 2.4, D65, 100 Nits, Video level throughout system). The Mac Studio is set to DM240 color profile in System Preferences. I also ran the SDI output of my Ultrastudio Mini 4k into the SDI input of the same DM240. I turned off "Use Mac display color profiles..." in DR Preferences. I used the split screen function on the monitor to display both inputs simultaneously. Aside from a very minor Hue offset (with the DVI input), which I easily corrected on the monitor, the images looked
IDENTICAL. If this same clip is then rendered out to QuickTime with 1-1-1 tags, the images looked
IDENTICAL to the SDI source and the DVI viewer source in DR. This means that all color management seems to be negated and you're left with the same signal that is being fed out of the Ultrastudio (which is not affected by apple color management).
If I turn on "Use Mac display color profiles..." in DR Preferences and set output color space to Rec709 [Scene] or Rec709 2.4 gamma, the image exhibits a darkening effect in the Viewer window. The SDI input and DVI input do not match now. Again the DVI input being darker. However, when rendering out the image, which applies 1-1-1 tags with Rec709 [Scene], the image matches SDI in Quicktime through DVI. But when rendering out Rec709 2.4 gamma which sets 1-2-1 tags, the image matches the Viewer window, but is darker than the SDI input. So, yes, the "Use Mac display color profiles..." definitely will match (actually the near blacks are very very slightly lifted in QuickTime, so not sure what that is), when viewing the same rendered file in Quicktime. But in relation to a known standard, I'm not sure what the monitor should be calibrated to, to exhibit a match with the known reference. Just doing some non scientific chimping, It appears that around 2.0 gamma gives it a similar appearance. But the colors still look a bit off.
To me, this NCLC stuff is overly complicated and not accurate at all. Maybe it's accurate to itself, but that doesn't matter in the real world and will have massive user error / failure, which is what we keep seeing over and over, in these forums. So that's why I tag with 1-1-1 and then let the user set up their system to reflect the known input. This allows people with correct calibration to see the correct image. I don't want people to be dealing with this automated and absurd color management system, that I've never seen work correctly, in the real world. But it, unfortunately, is the way it is.
The solution would be to have every monitor be calibrated with a 3D LUT based system. The tags would then simply be read by the software to say this file is to be viewed on a "Rec709 2.4 spec monitor" or a "DCI-P3 2.6 spec monitor". The software would then merely switch the monitor LUT automatically to the correct setting and voila the user would see exactly what the mastering house / DI saw. This would be absolutely amazing. The user would just need to recalibrate the system every so often and it would stay in spec. Sadly, it's what we hope the ColorSync system is doing, but it's really poorly implemented and not clearly defined. I've been researching this for weeks now, am technically savvy, I do this for a living professionally and I still am having trouble wrapping my mind around the concepts and why things aren't working. ICC basically sucks and was never designed for broadcast / video standards, but print. Even the white point is based around D50 and needs to be changed for video based profiles. So, again I agree with you, this is a mess. But thankfully, the broadcast system setup is the solution for now and it does work.
Also, does anyone actually know what "NCLC" stands for? I can't find it anywhere online.