Author Topic: Tonemapping - Plz Halp  (Read 115750 times)

2020-05-05, 18:16:07
Reply #210

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile
What's impressive about Fstorm is that outside you can see all the detail. With Corona is completely blown out.

2020-05-05, 18:23:56
Reply #211

lolec

  • Active Users
  • **
  • Posts: 179
    • View Profile
Not sure if anyone is still interested in this. If not, let me know.

Below I rendered a test with gray materials. I removed the glass. No DOF still.
I did Fstorm, Corona with cache and Corona brute force.
Corona with PT never cleared the noise, even after over 300 passes.



Can you post your Corona settings?

I find that increasing Highlight Compression would allow you to recover details outside, but others are posting different settings, curious about what settings you are using. 

Also, would you mind downloading the scene I uploaded and try it with the same Corona settings you are using for this ?


2020-05-05, 18:27:48
Reply #212

moriah

  • Active Users
  • **
  • Posts: 24
    • View Profile
Does Fstorm use GGX? Seems the BRDF is different compared to Corona, as there's more detail in the shadows.

2020-05-05, 18:36:24
Reply #213

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Thx for the test James, I'm dubitative with this conclusion tho. A wider colorspace only allows for a slightly better GI (corona already benefit from that). The lighting part you mention in Vray is the consequence of the tone mapping curve embedded in the ACES view transform.

Lupaz is right, the scene needs to use the same lighting (free hdri or constant color) and the same assets (I see the widows suddenly disappeared in the Fstorm scene). For those who are wondering, the textures are from the corona library. Also, matching the initial exposure is important to start on a similar basis imho.

I made a quick test in Vray with this HDRi : https://hdrihaven.com/hdri/?h=reichstag_1
camera : iso 1600 / F-number : 4 / shutter speed : 1/50s

Default sRGB :



ACES view trasnform with sRGB primaries (so no texture conversion) :



sRGB + Highlight compression + Handmade curve :





It took me 10s to build the curve by hand to my liking and it's really close to what ACES do and I still have control over the curve and the highlight compression.
« Last Edit: 2020-05-05, 22:38:59 by Fluss »

2020-05-05, 18:46:02
Reply #214

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Thx Lupaz, they are pretty close now. I guess tone mapping is activated Fstorm ?
« Last Edit: 2020-05-06, 14:47:49 by Fluss »

2020-05-05, 18:49:25
Reply #215

James Vella

  • Active Users
  • **
  • Posts: 542
    • View Profile
Thx for the test James, I'm dubitative with this conclusion tho. A wider colorspace only allows for a slightly better GI (corona already benefit from that). The lighting part you mention in Vray is the consequence of the tone mapping curve embedded in the ACES view transform.

Lupaz is right, the scene needs to use the same lighting (free hdri or constant color) and the same assets (I see the widows suddenly disappeared in the Fstorm scene). For those who are wondering, the textures are from the corona library. Also, matching the initial exposure is important to start on a similar basis imho.

Good to know, ill have to have a play with these settings tomorrow. Cheers!

2020-05-05, 18:55:49
Reply #216

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Thx for the test James, I'm dubitative with this conclusion tho. A wider colorspace only allows for a slightly better GI (corona already benefit from that). The lighting part you mention in Vray is the consequence of the tone mapping curve embedded in the ACES view transform.

Lupaz is right, the scene needs to use the same lighting (free hdri or constant color) and the same assets (I see the widows suddenly disappeared in the Fstorm scene). For those who are wondering, the textures are from the corona library. Also, matching the initial exposure is important to start on a similar basis imho.

Good to know, ill have to have a play with these settings tomorrow. Cheers!

I still find ACES a bit better (did not try to match the curve exactly tho) but that was just to show that the differences are not huge.

2020-05-05, 19:09:06
Reply #217

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile
Can you post your Corona settings?

I find that increasing Highlight Compression would allow you to recover details outside, but others are posting different settings, curious about what settings you are using. 

Also, would you mind downloading the scene I uploaded and try it with the same Corona settings you are using for this ?

Yes, definitely with highlight compression I'd start seeing

These were the settings before:


And this is with highlight compression. The interior starts to get dull.




2020-05-05, 19:11:28
Reply #218

Fluss

  • Active Users
  • **
  • Posts: 553
    • View Profile
Try to make an s-curve like this with corona curves:




2020-05-05, 19:19:20
Reply #219

MattiasD

  • Active Users
  • **
  • Posts: 15
    • View Profile
    • Key Visuals
Maybe that`s why a lot of people are asking for ACES implementation. It`s not the holy grail, but you wouldn`t be doing some custom curves, that just look close enough...
It would be amazing if the devs find time to implement it. I personally think it`s just a matter of time, since it`s getting, much like Dontnod`s PBR back in the day, more an industry standard.
And i`d really like to move forward and start complaining about other stuff :)

2020-05-05, 19:34:07
Reply #220

lolec

  • Active Users
  • **
  • Posts: 179
    • View Profile


Yes, definitely with highlight compression I'd start seeing


I always use HC in combination to contrast ( 3- 4 )  In a way, contrast is similar to making a S curve.

2020-05-05, 19:40:28
Reply #221

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile


Yes, definitely with highlight compression I'd start seeing


I always use HC in combination to contrast ( 3- 4 )  In a way, contrast is similar to making a S curve.

Yup. I do the exact same thing. HC at 2 and contrast at 3 is my default.

2020-05-05, 19:58:58
Reply #222

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
Yup. I do the exact same thing. HC at 2 and contrast at 3 is my default.
[/quote]


HC very quickly starts to eat the light dynamic in images, so they become sizer dull and dark.
I almost never increase HC over 1.1 / 1.2.
Instead you can crank up Filmic Highlights until you see details in overexposed areas, without sacrificing the overall dynamics of your image.

HC 1
Contrast 2-3
Filmic 0-0.9

Works in 99%




2020-05-05, 20:06:47
Reply #223

lupaz

  • Active Users
  • **
  • Posts: 951
    • View Profile
Yup. I do the exact same thing. HC at 2 and contrast at 3 is my default.


HC very quickly starts to eat the light dynamic in images, so they become sizer dull and dark.
I almost never increase HC over 1.1 / 1.2.
Instead you can crank up Filmic Highlights until you see details in overexposed areas, without sacrificing the overall dynamics of your image.

HC 1
Contrast 2-3
Filmic 0-0.9

Works in 99%
[/quote]

To me, filmic highlights makes everything even duller than HC. I guess it's a matter of taste.

2020-05-05, 20:11:32
Reply #224

Designerman77

  • Active Users
  • **
  • Posts: 507
    • View Profile
To me, filmic highlights makes everything even duller than HC. I guess it's a matter of taste.
[/quote]


Oh, that's really interesting.

In C4D / Corona, this behaves exactly the opposite.
I'm still using Corona 4.
Hopefully this HC / Filmic behavior I got used to is not gonna change in Cor. 5...