Page 1 of 1

Physical Sky + ACES

Posted: Mon Jun 27, 2022 6:34 am
by daros
Hi Dade, we have started to study Lux's code but we didn't understand what we need to do to make the physical sky work with ACES. Can you give us some tips please about how to intervene?

This is with no ACES
no ACES.JPG
no ACES.JPG (8.04 KiB) Viewed 9905 times
with ACES
ACES.JPG



In all other situations we get close to perfect results.

Re: Physical Sky + ACES

Posted: Tue Jun 28, 2022 7:11 am
by kintuX
daros wrote: Mon Jun 27, 2022 6:34 am Hi Dade, we have started to study Lux's code but we didn't understand what we need to do to make the physical sky work with ACES. Can you give us some tips please about how to intervene?

This is with no ACES
no ACES.JPG

with ACES
ACES.JPG




In all other situations we get close to perfect results.
Unsure but, if blue skews then something's wrong (either implementation or engine) :|
Have you made any methodical tests to "establish color correction" or performed comparisons to see/show what happens to image data?
What I can observe here & there (ACES implementation in FELIX.), all appears subjective.

I.E.
ACES OT on Corona White Light through Blue volume
(source: https://corona-renderer.com/comparer/MxEqUe)
AvsB.jpg

Re: Physical Sky + ACES

Posted: Tue Jun 28, 2022 1:35 pm
by daros
Hi Kintux, the following is one of the many many tests we did in order to tune ACES as good as we could to match linear.
No tricks, just applying the theory trying to avoid mistakes. The only thing i did was to adjust in post the exposure of the Linear in order to match roughly the greys of ACES.
If you have some specific ideas of other tests i could do in order to focus the problem please let me know, i will do them.

Lux Linear
color linear_1.jpg

Aces
color aces_1.jpg

Re: Physical Sky + ACES

Posted: Sun Jul 03, 2022 7:33 am
by Dade
The sky model doesn't take any "color" as input so there is no color to convert from. It is just a model coming from a paper and it works in linear space. The output is converted to the color space you picked as any other pixel.

However I have the feeling your problem may be related to sky (color) temperature setting: it can have a quite dramatic impact on the end result and you should try to play with that parameter.

Re: Physical Sky + ACES

Posted: Sun Jul 03, 2022 1:44 pm
by daros
Hi Dade, i tried to put the sky color at 12k kelvin and still the cyan dominance is strong.
i cannot get a deep blue sky.

ACES off
sky no aces_1.jpg

ACES on
sky ACES_1.jpg
however at cycles i see a similar behaviour... or at least time ago.
https://devtalk.blender.org/t/new-sky-t ... 28?page=27

Re: Physical Sky + ACES

Posted: Sun Jul 03, 2022 7:15 pm
by Sharlybg
I have tested Dade Idea thanks to a Bender addon Dev on Twiter.
( @Dade By the way people seem to not be able to create an account on the forum)
And Temperature setting seems to give nice result :
Sky_Luxcore.jpg
Here are the full test : viewtopic.php?p=34152#p34152

Re: Physical Sky + ACES

Posted: Sun Jul 03, 2022 9:33 pm
by Dade
Sharlybg wrote: Sun Jul 03, 2022 7:15 pm ( @Dade By the way people seem to not be able to create an account on the forum)
Russian spammers create 10-15 accounts, I have to cancel, every day...

Re: Physical Sky + ACES

Posted: Mon Jul 04, 2022 10:53 am
by daros
Ok... i think i have to make some more tests to see if our implementation is correct and then i come back to you. Thanks for the moment!

Re: Physical Sky + ACES

Posted: Wed Jul 13, 2022 3:18 pm
by daros
I didn't had the opportunity to make some more scientific tests or accurate comparisons. For the moment i followed the advice of sharlybg, to vary the sun color temperature:

This is the reference HDRI result which seems great to me. I have many blue sky HDR's like this and all give more or less the same result. In any case no color shift to cyan.
ACES View 01 1200x600_1.jpg

This one is physical sky at 6250K
no ACESView 01 1200x600-6250K_1.jpg

This one is physical sky at 5500K
no ACESView 01 1200x600-5500K_1.jpg

This one is physical sky at 4750K
no ACESView 01 1200x600-4750K_1.jpg