Author Topic: Tonemapping - Plz Halp  (Read 45853 times)

2020-04-23, 14:48:50
Reply #15

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
Regarding Lolek´s test: me personally I know those parameters in Corona and what effects they have, including LUTs. Using them all day long.
I guess 99,5% of us know those tricks.
However, a simple box or a living room with all furniture, etc... still seem to be two different things. :)
Who knows, maybe if you make all shaders super-duper physically correct, according to real life, it might get an extra kick of realism... ???
But who's time for such?

This morning I tried to look at good 3D images that stand out ( like those fron J. Lindqvist, Ferretti, etc. ) and have to say that a lot of the realism happens in the way they deal with the light
in therms of highlight peak, contrasts, etc. Those great images mostly have like a color layer that influences all colors... vs. too clean colors. And no crazy contrasts... since actually real world looks like this.
That's actually a thing one learns as an artistic painter at academies in the first semester. :)

2020-04-23, 17:31:45
Reply #16

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
Right now I was a bit stuck on lighting a bathroom that has no windows and some creme colored wall tiles... not the nicest thing for the light mood.... I hate such rooms. :)

Long story short: by mistake I unticked "tone mapping" while the KimAmland LUT was full on ( value 1 )... and voilà, suddenly the freaking grayish light in the bathroom changed to a much more natural look,
just like I see now in Lolek´s posting.

Dudes... this setting which Maru mentioned (camera settings on default + Amland LUT)... seriously, this should be standard Corona setting and the first basic tutorial.
As other users mentioned: a pity to fiddle around for years in the cam settings, the light, materials, etc - just to realize that camera on more-less base setting plus a good LUT is the better way to go.

Glad that this topic of light mapping came up here!





2020-04-23, 17:46:46
Reply #17

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
It seems a lot of people believe there is a magic button that makes everything look photoreal but that won't happen. We're 90% there with Dubcat's LUTs. The only missing bit is a slightly better handling of highlights. That's it. Materials, composition and lighting play a huge part too. In fact, the main issue here is that with Dubcat's LUTs, you're stuck with embedded curves. So a good tone mapping tool would let access to toe, shoulder and medium parts of the image, in an elegant manner (probably in log space like ACES is doing during the display transform). But there is no universal settings... So I disagree to make those settings default.

2020-04-23, 18:28:31
Reply #18

lolec

  • Active Users
  • **
  • Posts: 168
    • View Profile
I agree there is no "magic" setting. But Fstorm has a strong LUT as default and maybe that gives people the idea of it being so much better (it is not)

I think this is a separate discussion. After you've done every possible thing, Fstorm still handles tonemapping a little better. But I suspect what most people would notice is just enabling a photographic LUT.

That's why I think a single button with an attractive name which selects a photographic LUT enables bloom and maybe some other small tweaks would make a huge difference and unlock peoples brains to start exploring with those settings too.


2020-04-23, 19:53:54
Reply #19

agentdark45

  • Active Users
  • **
  • Posts: 570
    • View Profile
Thanks for taking the time to post up the Cornell box tests lolec.

While I agree that the Kim Amland LUT with minimal VFB fiddling produces a vastly super result than stock settings, the way Fstorm handles highlight peaks, the bottom curve of extreme darks and colours in general is still superior. It may not be as apparent in a simple Cornell box scene - start populating a real scene, with strong lighting certain areas and watch what happens to your highlights, strongly coloured materials and deep shadows; things start to break down and you end up chasing the dragon.

You can start to notice it in your Corona + Kim example 1; look at the exaggerated shadowing and black crush creeping in, as well as the front face of the small cube compared to the Fstorm tweaked example. Overall scene contrast is retained in the Fstorm tweaked example without exaggerated shadowing - it almost looks like light is behaving more realistically on the objects.
« Last Edit: 2020-04-23, 20:02:49 by agentdark45 »
Vray who?

2020-04-23, 20:48:35
Reply #20

cjwidd

  • Active Users
  • **
  • Posts: 924
    • View Profile
    • Artstation
@lolec that is a really compelling example, although I am reluctant to generalize too much from such a well-defined scene. That being said, the comparison does drive home your point about photographic LUTs.

Thank you for sharing!

2020-04-23, 23:04:00
Reply #21

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
Super important topic!
Maybe THE most important since long time. So thankful, all of you guys are participating with great input, ideas, opinions, etc.

I think many 3D artist would like to have a default setting that works like the human eye sees reality. Without lost details in highlight, nor with too blackish shadows, dull mid tones, etc.
And on top of this basic "eye-like" setting, more photo-typical behavior should be possible to - lets say - fade in with a slider. :))))

Of course, with enough grey cells, experience and try-and-error, one can always find a solution, also now.
But when clients demand great AND quick... well, then a render engine that simply does the magic out of the box would be the go to.

Today, for example, I lost hours with light setting... while a huge project has to be done as quickly as possible. And I know that my client will ask tomorrow for results.
So I'm sitting again now, at night and do work.
Costs nerves...


2020-04-24, 11:37:10
Reply #22

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Well, that's more related to how you market your work, to be honest. Making great images takes time. So great and quick is a bit wacky.

IMHO the topic is not to mimic the perceptual representation of a scene but a filmic representation of a scene that is probably more contrasty than what you would think due to dynamic range limitations. Terminology is really important.

Not to mention that pro photographers never use their images as is, they retouch them to look good, convey emotions, emphasizing some aspects of the scene, tweaking colors, bringing more light in some places etc etc.... So having a reference curve to lookdev is totally fine but that stops there for me. Hence the fact that a lot of us were pushing for ACES as it offers both a nice filmic preview AND a color management system that will allow you to deliver your work on a variety of media including HDR displays. That said, that's a fixed embedded curve on which you have no control (if you use default OCIO workflow) so it may disappoint a lot of people.

Not to mention that such filmic curves compress highlights so much that you will have to increase energy levels of your scene which increase realism but will also have an impact on render times (more overblown areas). Unreal engine implementation fakes it by applying a 1.45 gain before the post-process because they did not want artists to be bothered by the exposure loss (was a bad choice imho).

2020-04-24, 13:44:36
Reply #23

Juraj

  • Active Users
  • **
  • Posts: 4305
    • View Profile
    • studio website

Not to mention that such filmic curves compress highlights so much that you will have to increase energy levels of your scene which increase realism but will also have an impact on render times (more overblown areas). Unreal engine implementation fakes it by applying a 1.45 gain before the post-process because they did not want artists to be bothered by the exposure loss (was a bad choice imho).

This doesn't make sense. Exposure increase is identical to increasing light source intensity. One stop of Light is double the light (since one is exponential and other linear). That's not opinion, that's fact.
None of this affect render times negatively per se, that's the purpose of MSI to control.

These threads are just getting weirder and weirder.

(Great tests Lolec! btw).
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2020-04-24, 15:04:02
Reply #24

mantaskava

  • Active Users
  • **
  • Posts: 72
    • View Profile

Not to mention that such filmic curves compress highlights so much that you will have to increase energy levels of your scene which increase realism but will also have an impact on render times (more overblown areas). Unreal engine implementation fakes it by applying a 1.45 gain before the post-process because they did not want artists to be bothered by the exposure loss (was a bad choice imho).

This doesn't make sense. Exposure increase is identical to increasing light source intensity. One stop of Light is double the light (since one is exponential and other linear). That's not opinion, that's fact.
None of this affect render times negatively per se, that's the purpose of MSI to control.

These threads are just getting weirder and weirder.

(Great tests Lolec! btw).

So here's a situation/scene - a single room with a single light in it. Now there's two ways to brighten it up, either you brighten your light or increase your EV/Exposure. Do I understand correctly that there's totally no difference which method you choose? Even if you would brighten your light to some crazy values (like 20 000 of whatever), it would still be okay? By okay I mean materials would still respond to light as expected etc.

2020-04-24, 15:46:21
Reply #25

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile

This doesn't make sense. Exposure increase is identical to increasing light source intensity. One stop of Light is double the light (since one is exponential and other linear). That's not opinion, that's fact.


Did I said anything otherwise? What's the point here?

None of this affect render times negatively per se, that's the purpose of MSI to control.

Well, this may not apply to corona sampling method but definitely would with Vray bucket mode. Renderer has hard time sampling bright areas. Here are two renders with the exact same settings, just different display transform (so the linear render behind both images is exactly the same):

sRGB


ACES


If you wanna match ACES perceptual light intensities to the sRGB one, you'll have to increase the exposure. So the linear render will expose more overblown areas with which the renderer struggle to deal with. Of course we're not talking tremendous amounts but still may count in animations. That what I experienced in my tests.
But as I said earlier in this post, it may indeed not apply to corona as it is a different sampling method. Maybe I took a bit of a shortcut here, I admit. It still is not devoid of foundation. I'm pretty sure you did encounter that Juraj, as you were a Vray user. Remember the last bucket of death ?

source refering to the UE tonemapping gain : https://acescentral.com/t/creating-an-ocio-view-transform-to-match-default-unreal-4-aces-viewport-tonemapper/2437
                                                                  https://acescentral.com/t/acescg-for-animation-feature-and-further-questions/1545/21

"Keep in mind that Unreal Engine 4 ACES fitting applies a 1.45 gain, i.e., it is over-exposed compared to ACES reference implementation. This is an aesthetic choice done so that artists at Epic Games had an easier transition to ACES compared to a simple Gamma 2.2/sRGB OETF or a tonemapper keeping brightness in the ballpark."

So to conclude, ACES tonemapping allows you to bring more energy into your scene which makes quite a difference. But is it not the point of a tonemapper ?
« Last Edit: 2020-04-24, 17:07:56 by Fluss »

2020-04-25, 00:14:36
Reply #26

cjwidd

  • Active Users
  • **
  • Posts: 924
    • View Profile
    • Artstation
These threads are just getting weirder and weirder.

Hopefully you mean these *comments* are getting weirder and weirder - the thread is a pretty honest request for help on a diverse topic.

Those great images mostly have like a color layer that influences all colors... vs. too clean colors. And no crazy contrasts... since actually real world looks like this.

@Designerman77 can you elaborate on what is meant by "too clean colors"?

the way Fstorm handles highlight peaks, the bottom curve of extreme darks and colours in general is still superior.

F-storm always comes up as the comparison group on Corona threads when tonemapping is concerned, and what is described here^ is often the punchline of those conversations.

Is there really any disagreement about this, and if not, how is that reflected in the tonemapping rework that appears on the roadmap?
« Last Edit: 2020-04-25, 12:47:55 by cjwidd »

2020-04-25, 12:20:23
Reply #27

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Well if you guys consider those comments weird then I'm gonna stop posting from now. I always try to add valuable insights and I always try to document my posts with links and tests, as much as possible. And I spend time for it as english is not my mother language. I also often mention that all the stuff I write is the reflect of my personal opinion and interpretation of things. I'm fine with contradiction as long as there is some argumentation to go with it. So I'm gonna stay far from those technical threads from now, I'm tired of this TBH.

2020-04-25, 12:29:20
Reply #28

cjwidd

  • Active Users
  • **
  • Posts: 924
    • View Profile
    • Artstation
@Fluss I want to be clear that my comment was not targeting you and I appreciate you taking the time to share your insights on the forum +1
« Last Edit: 2020-04-25, 12:45:32 by cjwidd »

2020-04-26, 00:29:02
Reply #29

Jpjapers

  • Active Users
  • **
  • Posts: 1443
    • View Profile
oh god this is going to turn into one of those threads isnt it haha