AgX in Corona

[zhao liang]

Thank you for your answers!

[piotrus3333]

More of AgX by Troy Sobotka - this time his Resolve version of it (quite different) married with a film look from Arri:
https://forum.corona-renderer.com/index.php?topic=41947.msg223544#msg223544

note: it does require a string option to turn off VFB's sRGB view lut so might be a bit more cumbersome to use.

[muoto]

But when importing exr's from corona to fusion, there is a extra gamma applied (1.2?) because the image in fusion is much darker than in VFB. And then you need to specify the float gamma thing to have the same result in vfb and fusion ?

[piotrus3333]

But when importing exr's from corona to fusion, there is a extra gamma applied (1.2?) because the image in fusion is much darker than in VFB. And then you need to specify the float gamma thing to have the same result in vfb and fusion ?

AgX luts for Corona from this thread are made for default Corona behaviour regarding gamma transforms in VFB. by default VFB has hidden 2.2 transform after the TONE MAPPING stack and effects. it also has hidden gamma transforms in every LUT operator.
if you are using my AgX luts there is no need for string option changing VFB's gamma behaviour ("float gamma thing").

[pokoy]

But when importing exr's from corona to fusion, there is a extra gamma applied (1.2?) because the image in fusion is much darker than in VFB. And then you need to specify the float gamma thing to have the same result in vfb and fusion ?

AgX luts for Corona from this thread are made for default Corona behaviour regarding gamma transforms in VFB. by default VFB has hidden 2.2 transform after the TONE MAPPING stack and effects. it also has hidden gamma transforms in every LUT operator.
if you are using my AgX luts there is no need for string option changing VFB's gamma behaviour ("float gamma thing").

Wait, so for any LUTs from outside Corona install we need either the string option or a gamma operator @0.45 before the LUT? I remember trying some LUTs from other sources but gave up thinking they were not the right type.

[piotrus3333]

Wait, so for any LUTs from outside Corona install we need either the string option or a gamma operator @0.45 before the LUT? I remember trying some LUTs from other sources but gave up thinking they were not the right type.

I do not know the story behind luts from Corona installer but, in general - yes - to have a lut do what it is supposed to do in Corona VFB you need gamma transform operator (1/2.2).
I guess the story why it all works like that is long forgotten as the support was unable to point me to the simple solution - the string option for adjusting VFB gamma. I dug it out from an old 2015 post.

[piotrus3333]

ok, so at least we know that it used to work as expected for a brief moment:

[muoto]

would anyone have a nuke layout that would work creating LUT's that would look similar in Nuke as in Corona Vfb ?

[mienda]

Hello, I just saw this video on YouTube which deals with a so-called new tone mapping for Blender "he Khronos PBR Neutral Tone Mapper" (

)  which seems to have the advantages of AGX without its disadvantages (lack of contrast and saturation depending on everyone's tastes). I have the impression that the combo of AGX MATRIX + AGX PUNCHY (+ contrast if necessary) is close compared to the combination AGX BASE + AGX KRAKEN + AGX DEFAULT CONTRAST.
I'm really uneducated about these issues so I was wondering what you thought about it.

 Thanks in advance

[piotrus3333]

AgX in the form of luts from this thread is not the same AgX you have in Blender. There are three different versions of AgX - the one I uploaded is based on the original idea from Sobotka. it does not shift hues like the Blender 4 version. it is also made with BT709 primaries (what you would colloquially call srgb primaries) in mind.

agx matrix + punchy is, well, more punchy than agx matrix + kraken + default contrast. default contrast is called base look in original AgX ocio configuration. you can replace default contrast lut with your own curve or add colour grading lut on top of Kraken. punchy look if I remember correctly is just gamma adjustment - just a simple sample look, nothing to stick with. so this AgX basically will require your personal input in regards to final contrast.

it is possible than some other, preset based, display transform would be better suited for use in Corona: for example RED display transform consists of 16 combinations of 4 levels of contrast and 4 different highlight roll off strengths.

Khronos PBR Neutral Tone Mapper - I think they are trying to solve issues with colour reproduction for some quite specific cases. does any image in the video look natural to you? instead of rendering in wider gamut space and using some tried and tested colour science (from camera manufacturers for example) they are fiddling with display transform alone. BT709 colour space is too tiny to describe all real life colours - look at professional cameras: at least BT2020.
EDIT: ok, so this is indeed designed specifically for e-commerce. very different beast, very different needs. not for us.

[mienda]

 
Thank you very much for your answer Piotrus. This clarifies these tone mapping things that are still confusing for me

[piotrus3333]

Another example of using AgX:
Resolve AgX version (DCTL) married to look from famous ARRI K1S1 lut.

[alexyork]

Hi Marcin, folks/devs,

We are typically using your fantastic AGX ToneMapping instead of ACES for almost everything these days, from this method, as it seems to have a much more pleasant and predicable rolloff for the highlights while preserving contrast and colour. The issue we're facing is that it doesn't seem to work properly with ToneMapControl when you set it to not affect the map. It seems to affect the exposure and crush certain tones and mess around with the contrast. I see there's a warning for this on the corona website "While the Corona Tonemap Control Map works fine most of the time, there are some special cases where it may fail, for example: using some LUTs, using extreme tone mapping operator values, setting the Contrast value to zero."

- Is there any workaround for this to get 1:1 result with the underlying map, like we get with ACES workflow?
- Is there a simplified single-LUT version of this we can use, which may or may not help?

Some notes:
- corona 11 HF 2, max 2024
- base map is an SRGB photo loaded into coronabitmaploader, zero colourcorrection, just the pure map
- it's then piped into tonemapcontrol with all the checkboxes disabled
- using AGX workflow method from here: https://forum.corona-renderer.com/index.php?topic=40952.0 - it results in 2 x LUTS being loaded into the coronacamera tonemapoverride. It's the second (lower) of the two LUTs that's breaking things (or rather being broken by the tonemapcontrol). And I believe we have the correct Logarithmic settings
- contrast is at 1.0, or 1.5 it doesn't matter (but it's not at 0.0 FYI, which we know can break things)
- we're aware of the IR tone mapping bug where you need to re-start the IR each time you affect things like exposure, tone mapping, LUTS etc. which we've done, so this isn't the issue (although devs, any news on a fix for this? it's very annoying and easy to forget!). It happens also in production rendering.

Example below:



Thanks for any help!

[piotrus3333]

@alexyork

skip the Corona Tonemap Control and just transform the backplate into your rendering space, linear rec709 in this case.
for this you need to reverse the AgX Punchy display transform. lin rec709 backplate will get back to AgX Punchy via frame buffer corrections. if you need to colour correct the backplate in Max do it in lin rec709 so its nice and tidy.

you need the original AgX ocio config from Troy Sobotka: https://github.com/sobotka/AgX

attached screenshot shows conversion done with VRayOCIO node, you can do this with Affinity/Photoshop/whatever else. 3dsMax 2024 setup with this AgX OCIO file should be able to convert it too.
there will be slight differences as luts I made ale limited in precision by Corona's cineon log conversion in lut layer.

[piotrus3333]

you can actually also try these luts. made for Vantage so you will always end up with transfer function difference srgb vs 2.2.
I also never tested Corona's linear option in lut layer so no guaranties.
instead of log - use Reinhard curve (h compression 999 in Corona) - should be more precise compared to old double lut setup.