here’s why : AgX in Blender 4.0. It’s coming, are you ready? - YouTube
I couldn’t agree more!
AGX is only a tonemapper, so it should be somewhat easy to implement into Vray
It’s a single 1d LUT with a dedicated OCIO config.
It seems to work fine as it is, it only expects a linear BT.709 colorspace as input, and offers three view transform options.
It has other little things but i am unsure as to what you’d like us to develop.
Thanks for that link. That was an excellently informative piece with clear conclusions.
The explanation of everything is probably the best and most understandable I’ve come across to date.
Hi Lele
You guys developed Filmic into Vray. Why not do the same with AGX?
I know it is as “simple” as loading the OCIO file and make sure all the settings are according to your image, but why not just implement AGX into Vray itself from the beginning, so users have easier access to the tonemapper?
I would imaging that it would be a layer that you load into the VFB, just like you do with Filmic at the moment.
Filmic was a game changer when it first came out, but it took Chaos ages to implement it into the software.
Blender pretty much implemented it from day 1, and everybody loved it.
Now AGX is on the rise and looks more capable than Filmic.
Optimally everybody would use ACES, but it is too much a hassle to set up correctly and you have to double check all the bitmaps in the Mat Editor and make sure all the color settings are correct.
Of course all of this can be avoided by naming all the texture files correctly, but that is a HUGE task that very few studios have time for.
AGX is like the perfect middle solution, since it’s a Tonemapper and not a color system. It delivers the same kind of colors as ACES, just without all the chaotic bitmap management.
Hope it makes sense.
I think it needs the OCIO config to work, the LUT (i.e. the tonemapping part) without the transforms the config holds wouldn’t work.
It’s also not editable, one can only pick and choose, and at most blend, as it’s not procedural, but deterministic (i.e. it doesn’t have knobs to tweak, unlike, f.e., Filmic.).
As such, it’s highly unlikely it could work besides the rest of the (wholly procedural) tonemappers.
Given it’s exclusive with the use of ACES, the provided ocio config and LUT seem to work just fine, don’t you think?
The devs are reading this, rest assured, it’s just that it’s not obvious what should be done on our side, besides allowing you to use it in its current form.
Interesting!
I see your point, wouldn’t make sense if you can’t tweak the result.
Just to be clear, wouldn’t i need to convert the color profile from Linear --> Linear BT.709 using the ACES OCIO, and then convert again from BT.709 --> sRGB in order for AGX to be presented correctly, or can i use the setup like you showed in the screenshot?
The two linear options are one and the same that i know of:the primaries are identical, at the very least.
There may be fairly minor differences with the sRGB color matrix, but they wouldn’t likely be noticeable, and surely not math-breaking. (edit: there are slight gamma differences, a 2.2 for sRGB vs. a 2.0 for Bt. 709.)
The setup i used, with sRGB primaries for the renderer worked pretty much as expected (for a 1D LUT) and as shown in the samples: it worked great in specific parts (those that somehow matched the LUT’s expected ranges), and was pretty bad in others (those out of the expected gamut).
see here:
sRGB
Rec. 709
AgX
AgX Punchy
Agx Golden:
there is bit more to AgX:
- srgb input goes through colour matrix (so with a bit of imagination you can consider it “AgX colorspace”)
- that gets encoded into log space called Kraken (pretty ugly log for my taste, range 0.000175… to 16.291742…)
- only then the 1Dlut in the config puts this log onto a display - AgX Base, a very neutral look, lacking quite a bit of contrast if you’re coming from aces odts but providing a great, well, Base.
you can play with AgX and ACEScg with this setup:
Render with ACEScg and experiment with AgX on top. - Chaos Forums
if you do not want to tinker with adjusting primaries you can find some AgX ocio configs that added ACES colour spaces on github.
and for just comparing aces odt with AgX - use this config with added “aces contrast lut”:
AgX setup - V-Ray for Maya :: General - Chaos Forum
With this setup i get *far* too saturated results, making punchy and golden unusable, and the base AgX itself well too harsh. I also get a metric ton of red cast.
Admittedly, display Native looks closer to the source with your setup.
We’ll wait until this is fleshed out properly, and figure out if it’s worth pursuing specifically then.
With this setup, is it still necessary to convert all the bitmap primaries to ACEScg for everything to behave “correctly”?
there is nothing to pursue here. with ocio support there is no need to implement anything. load your config and it’s done.
”too saturated” and “red cast” sounds like srgb input. my VRay VFB layer setup expects acescg. the composite part does the conversion from acescg to lin srgb
(edit: plus AgX matrix)
, curves creates Keaken Log and ocio on top converts Kraken to whatever you need.
Agh, duh me, you’re very right.
sure, regular “acescg in VRay” textures treatment.
one very important note: all the videos on yt comparing aces with agx (including this one) for reasons unknown use aces 1.2 and not 1.3 (from 2021). aces’ “hot blue turning purple” is not an issue any more. if you see this shown as an issue with aces you can assume missinformation or at least incompetence.
Allright, i made ACEScg renders of the same scene, and using my approach now results to nigh identical results to yours. 
First image is the standard sRGB LWF render.
On its right, a standard SDR ACEScg render of the same scene (max 2024 defaults, to be clear.)
The line below shows the four look options using the expensive transform from ACEScg to Kraken Log and back.
The bottom line shows the same four options, but simply feeding the OCIO setup a linear input.
the slightly bigger image is here.
This should be a live ‘battle’ type thing at every VrayCon…with a shark tank for the loser…tickets please 
Image from:
Testing_Imagery/mery_lightSaber_lin_srgb.exr at main · sobotka/Testing_Imagery · GitHub
this should show some differences between aces odt and AgX odt:
edit:
ACES 1.3 with default gamut compress (built in transform from Resolve)
AgX - aded lut to roughly match contrast of aces odts AgX setup - V-Ray for Maya :: General - Chaos Forum
ocio config for testing AgX with some ACES mixed in (improves both acescg and linsrgb input) :
https://forums.chaos.com/forum/chaos-common/chaos-common-public/1191941-render-with-acescg-and-experiment-with-agx-on-top?p=1195589#post1195589