The result looks off.
Again it’s OCIO’s limitation, OCIO only offers View and Look binary. It’s very limiting.
It’s the one that forms the image. You draw your apple on canvas, of course the paint is not the exact color of the apple’s spectral reflectance.
Yes u r really rightt, that ''desaturate filmic ‘’ is incredible it turn filmic to more AGX look with slider thanks the linear desaturation on light values, the things u do on that node is really needed to be added for user control, u almost read my mind about which controls to expose.
Yes after realtime compositor, ur approach will be the best
The lack of control I’m talking about is only related to using it in the render panel, because it will continually change intentional grading. It’s fine in the way I’ve shown above.
Do you intend to include your argx in the convert colourspace node’s options? Otherwise for me it would present all the same issues.
Err, the base grey variable. It’s part of the AgX setup. You can’t modify that from within Blender (as yet). I’m saying it’s a thing that could be tweaked about AgX
Sure, I know I mean how can I tweak in the files, not in blender? in config? or is it in a certain file in looks? I ll try to find it if it use the same terminology
like something here or?
The options in the render panel, specifically view and look, is by OCIO’s limiting design. We can’t do much other than stuff in things into packed views and looks.
Yes but there will be several of them corresponding to the types of display device we support, like display P3.
already is included there. In fact, so is Filmic. In AgX as it currently stands, Filmic only appears in the Color Conversion nodes, not in the Color Management panel
Oh cool. I’ll download the latest build then. I’m on 3.4.0 alpha, but it must have been added recently as it’s not there for me:
Oh you need to download the AgX OCIO config, not the latest Blender version. AgX is not yet in Blender by default.
Btw @Eary why is the process space for Greyscale BT.2020 I-D65 rather than BT.2020 I-E?
I am using the 2012 luminance of BT.2020 Troy posted, what he posted was I-D65 so I figured I would just do that. Doesn’t make that much of a difference anyways.
ah I see. I think I know how to do the I-E version of that, but don’t have time for that right now. Can provide them soon
For those who are trying to test AgX, make sure you do it properly (by moving both OCIO config AND LUTs to blender color management folder, I used to only copy the OCIO file and I would end up with some funky results, also keep the old color management files somewhere in case you want to revert back to default).
you don’t have to delete. In fact, I think you should keep the Filmic folder if you want Filmic to be available in Convert Colorspace nodes (it will not be accessible in the Colormanagement menu)
and you can easily back up the other two if you think you might want to switch back and forth occasionally, and you don’t want to figure out how to combine configs. - Simply chuck “luts” and config.ocio into a new folder or something.
This is an interesting idea. I just tried implementing a slightly more sophisticated version of that with AgX.
The results are not great. But maybe something like this could work in principle, with some more thought.
This is the color sweep, assumed to be in Rec.709 I-D65. This isn’t the latest version of AgX though.
Rather than relying on Blender’s Hue node, I “corrected” that hue in the moment plane, which is more relevant to perception.
Display Native:
For some reason, correcting the hue in this way kinda breaks blue and, to a lesser extent, red, so it won’t be directly applicable in this way. But I don’t think this is necessarily a dead end.
Also, if you for some reason don’t want the filmic whiteout, I guess this is as far as you can possibly get away from the Notorious Six that way
It’s now the notorious 12ish instead. They actually are all distinct even at EV15, but for most of them you can’t really tell
Just to prove that, at least for now, this approach isn’t working, here some test imagery:
And the Line of Purple
with regular AgX
“moment-Hue corrected”
Matas actually looks almost ok imo (though quite different from normal), but its noise floor is raised extremely and there is definitely some clipping.
xmas Red also exhibits a lot of clipping
Blackbody is interesting because in some ways this actually is “better” - the hues are really nicely preserved for quite a bit longer. But the luminosities are off, and the purple end of the spectrum is definitely just broken.
I wonder how much of that is “pure AgX” and how much might actually just be the saturation compression stages that now potentially don’t quite act right as the colors in question are overwritten by different hues
If you place the hue correction after converting back to linear you get a nicer result. And the desaturation can be adjusted by the user adjusting the multiply between the rgb to bw and the saturation.
https://www.dropbox.com/s/hji372ypekbm81d/comp%20transform%20improved%20hue%20correction.blend?dl=1
You can pump the levels up crazy high and still retain the correct hue:
Or alternatively use lightness for the desaturation if you don’t want it to be desaturated based on perceived brightness:
Lightness (not as good)
Perceived:
Reminder, the research about Lightness and Brighness is still on going and all models available in Blender compositor that claim to achieve “perceieved brightness” don’t actually work.
Hmm how do you quote a video on this forum?
Here is the actual version 12:
Thank youuu for letting know about the new version. Nice to is its BT.709 again.
