Author Topic: ACEScg  (Read 9699 times)

2020-03-02, 22:58:00

maurus

  • Active Users
  • **
  • Posts: 8
  • 42.
    • View Profile
    • Maurus's Insta
Hey everybody!

How about ACES (or ACEScg) support for Corona? Linear kinda works though it would be nice to be able to get the maximum out of my renders with ACEScg in my ACEScct workflow.

Cheers
Maurus
"The ships hung in the sky in exactly the same way that bricks don't."  -Douglas Adams

2020-03-03, 09:30:02
Reply #1

Nejc Kilar

  • Corona Team
  • Active Users
  • ****
  • Posts: 1058
    • View Profile
    • My personal website
+1 :)

2020-03-03, 10:22:29
Reply #2

Jpjapers

  • Active Users
  • **
  • Posts: 1473
    • View Profile
This has been discussed at length before. Heres a good thread

https://forum.corona-renderer.com/index.php?topic=14958.0

2020-03-03, 21:44:03
Reply #3

SharkNinja

  • Active Users
  • **
  • Posts: 8
    • View Profile
Wow!  That thread is a BEAST!  I read the majority of it and to be honest couldn't tell if there was any resolution to it.  Most people were saying there are too many variables involved to successfully implement ACEScg into Corona?  At least that's what I took from it.  How is Arnold doing it so successfully then?

+1 ACEScg

2020-03-03, 22:10:58
Reply #4

davemahi

  • Active Users
  • **
  • Posts: 153
    • View Profile
    • iamstatic

2020-03-05, 02:23:06
Reply #5

agentdark45

  • Active Users
  • **
  • Posts: 577
    • View Profile
some video info...

https://www.youtube.com/watch?v=4cCQ3pTRfmU&amp=&feature=emb_title

Now this is what I'm talking about. The comparison between ACES and SRGB when pushing the light/exposure/saturation is exactly what we need in Corona. No more ugly linear-esqe burnouts and colour weirdness (and conversely ultra clamped muddy grey renders to get things under control), but instead actual filmic response like Fstorm.

This should be one of the top priorities for the dev team imo as it's a fundamental part of the output/realism of the engine.
Vray who?

2020-03-05, 13:55:38
Reply #6

Jpjapers

  • Active Users
  • **
  • Posts: 1473
    • View Profile
some video info...

https://www.youtube.com/watch?v=4cCQ3pTRfmU&amp=&feature=emb_title

Now this is what I'm talking about. The comparison between ACES and SRGB when pushing the light/exposure/saturation is exactly what we need in Corona. No more ugly linear-esqe burnouts and colour weirdness (and conversely ultra clamped muddy grey renders to get things under control), but instead actual filmic response like Fstorm.

This should be one of the top priorities for the dev team imo as it's a fundamental part of the output/realism of the engine.

I believe the filmic response youre looking for would be covered with the 'rework tonemapping' feature request.

2020-03-06, 11:33:45
Reply #7

JViz

  • Active Users
  • **
  • Posts: 139
    • View Profile
    • Behance
some video info...

https://www.youtube.com/watch?v=4cCQ3pTRfmU&amp=&feature=emb_title

Now this is what I'm talking about. The comparison between ACES and SRGB when pushing the light/exposure/saturation is exactly what we need in Corona. No more ugly linear-esqe burnouts and colour weirdness (and conversely ultra clamped muddy grey renders to get things under control), but instead actual filmic response like Fstorm.

This should be one of the top priorities for the dev team imo as it's a fundamental part of the output/realism of the engine.

I believe the filmic response youre looking for would be covered with the 'rework tonemapping' feature request.

there are two response curves as far as I know, the digital response curve which is very close to the human eye, and the filmic response curve which is more appealing although doesn't show the image as it would to a human eye in reality, it just looks nice.

and we have linear of course in corona, which needs to be processed to emulate a response curve then graded.

what we miss of course is the photographic/human eye response curve, we're essentially getting raw data that needs processing, it doesn't have to be this way. we're at a point where we zeroed in on the issue we're all facing. response curve! in Fstorm even if you have no idea what you are doing, if all the materials are wrong IOR glossiness diffuse you name it, the image still looks more realistic than corona simply because of the response curve that affects the scene in a way that makes it look real.

you can apply your curve at least to the luminosity of the image but then again color shifts will mess up the image and fine tuning is needed to get past the CGI fakeness barrier
Although a purist, my work is anything but.
https://www.behance.net/ImageInnate

2020-03-06, 14:14:00
Reply #8

Jpjapers

  • Active Users
  • **
  • Posts: 1473
    • View Profile


<SNIP>

there are two response curves as far as I know, the digital response curve which is very close to the human eye, and the filmic response curve which is more appealing although doesn't show the image as it would to a human eye in reality, it just looks nice.

and we have linear of course in corona, which needs to be processed to emulate a response curve then graded.

what we miss of course is the photographic/human eye response curve, we're essentially getting raw data that needs processing, it doesn't have to be this way. we're at a point where we zeroed in on the issue we're all facing. response curve! in Fstorm even if you have no idea what you are doing, if all the materials are wrong IOR glossiness diffuse you name it, the image still looks more realistic than corona simply because of the response curve that affects the scene in a way that makes it look real.

you can apply your curve at least to the luminosity of the image but then again color shifts will mess up the image and fine tuning is needed to get past the CGI fakeness barrier

The reworked tonemapping feature request came about because of the srgb/linear megathread and was intended (if i remember correctly) do mimic the DSLR response curve.
Dubcat has a set of settings that emulate ACES in corona which are...



Whenever people talk about ACES there seems to be alot of different topics get rolled in to one. Usually the whole sRGB/Linear/Filmic response stuff.

2020-03-06, 15:10:47
Reply #9

agentdark45

  • Active Users
  • **
  • Posts: 577
    • View Profile
^Those VFB settings are pretty decent, and are a good starting point but as JViz noted, colours can get extremely weird and don't react how they should IRL.
Vray who?

2020-03-06, 15:13:17
Reply #10

Jpjapers

  • Active Users
  • **
  • Posts: 1473
    • View Profile
^Those VFB settings are pretty decent, and are a good starting point but as JViz noted, colours can get extremely weird and don't react how they should IRL.

True but thats not what ACES is for. ACES is for preserving all of the colour data and pushing it into a universal colour space.
You have to apply a LUT (or RRT in the aces pipeline). All ACES colourspace does is standardise the input to the pipeline so you can work in a more standardised way.
Those settings as far as im aware bring coronas standard output into as measurably close to ACES colourspace as dubcat could get.
« Last Edit: 2020-03-06, 15:21:02 by Jpjapers »

2020-03-08, 14:17:09
Reply #11

Nejc Kilar

  • Corona Team
  • Active Users
  • ****
  • Posts: 1058
    • View Profile
    • My personal website
I might be getting it wrong but afaik ACES can help us control the colors and the highlights better because of the additional color data? So that's where you get that filmic look and the right color reproduction.

Could be very much wrong though :D

2020-03-30, 15:55:51
Reply #12

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 10956
  • Marcin
    • View Profile
Already reported; duplicate.

(Internal ID=275504302)

2021-10-01, 13:29:36
Reply #13

olivierc

  • Active Users
  • **
  • Posts: 5
    • View Profile
Hi Guys,

So what's up with ACES implemented in Corona 7?

Vray does it.. :/

Cheers,
Olivier