Author Topic: dubcats secret little hideout  (Read 93971 times)

2018-01-19, 22:29:12
Reply #120

zuliban

  • Active Users
  • **
  • Posts: 56
    • View Profile
I gave up along time ago in renders VFB i just use the lineal image as open exr and do the rest in nuke.
Still renders need small tech improvements like a correct DOF, cat eyes effect and obstruction  refraction, dof where you can see some small rainbow in some bokeh shapes and some hairs obstruction from grass in other shapes, a more correct material that preserves energy on bump and so on it.
It also would be nice to have this correct tone mapping that can mimics real world cameras and old school films.

And of course caustics... i feel nobody is looking at this anymore (corona roadmap have them)they really brings a image to life when it has enough glass or reflective surfaces.
I also have tested fstorm and i agree it feels like a real camera inside a render and create very realistic fast renders, but still i don't like the gpu low memory limitation make them really unusable for big textures or closeups maybe in some years we can have some nice 64gb cards and if miners are cool a good price range.

I also have seen some tech from Arion render and it has alot of nice features that other renders don't have like the dof obstruction and better materials i wonder why people don't use that render?
i tested it when it introduced bloom and glare before any other render then they deteached it to arionfx plugin.

2018-01-22, 16:32:29
Reply #121

oddvisionary

  • Active Users
  • **
  • Posts: 194
    • View Profile
    • Behance | Artstation : artstation.com/oddvisionary
I gave up along time ago in renders VFB i just use the lineal image as open exr and do the rest in nuke.
Still renders need small tech improvements like a correct DOF, cat eyes effect and obstruction  refraction, dof where you can see some small rainbow in some bokeh shapes and some hairs obstruction from grass in other shapes, a more correct material that preserves energy on bump and so on it.
It also would be nice to have this correct tone mapping that can mimics real world cameras and old school films.

And of course caustics... i feel nobody is looking at this anymore (corona roadmap have them)they really brings a image to life when it has enough glass or reflective surfaces.
I also have tested fstorm and i agree it feels like a real camera inside a render and create very realistic fast renders, but still i don't like the gpu low memory limitation make them really unusable for big textures or closeups maybe in some years we can have some nice 64gb cards and if miners are cool a good price range.

I also have seen some tech from Arion render and it has alot of nice features that other renders don't have like the dof obstruction and better materials i wonder why people don't use that render?
i tested it when it introduced bloom and glare before any other render then they deteached it to arionfx plugin.

You should try Indigo Render for caustics. The best for now, and they use MTL (Metropolis Light Transport) algorithm calculation, that are not implemented to Corona, which is why Corona caustics are nice but not the best.

You can get better caustics when using experimental settings : Bidir/VCM (if I remember correctly, I only tried once to use it for tests).
« Last Edit: 2018-09-23, 03:20:25 by oddvisionary »
Freelance Post-Prod / Lighting & Look Dev 3D Generalist | VFX Designer | Sound Effect Recordist & Sound Designer

Corona Discord server : https://discord.gg/2uxq8EA

2018-01-23, 00:20:50
Reply #122

snakebox

  • Active Users
  • **
  • Posts: 487
    • View Profile
    • Snakebox Media
Quote
You should try Indigo Render for caustics. The best for now, and they use MTL (Metropolis Light Transport) algorithm calculation, that are not implemented to Corona, which is why Corona caustics are nice but not the best.

I don't mean to hijack the topic, but I was pleasantly surprised by Redshifts caustics.

small test to demonstrate:
https://vimeo.com/239298771

It's still a post process like in vray I believe, but I found them much easier to use and also liked the result better. Caustics in Corona does not make sense to me, doesn't give me what I expect nor do I feel like I have any control. Hopefully that changes one day.

2018-01-23, 10:57:56
Reply #123

romullus

  • Global Moderator
  • Active Users
  • ****
  • Posts: 5724
  • Let's move this topic, shall we?
    • View Profile
    • My Models
I'm not Corona Team member. Everything i say, is my personal opinion only.

2018-01-26, 01:47:22
Reply #124

dubcat

  • Active Users
  • **
  • Posts: 456
  • ฅ^•ﻌ•^ฅ meow
    • View Profile
This is from an older post I made on the forum somewhere, but I want to consolidate all the important stuff in this thread.

This is how Corona behaves today.



This is default fstorm.



These are fstorm with adjusted parameters.



fStorm at 0.99 is the same as current Corona



I just want to state again that I'm not trying to provoke, I don't like to make 1:1 comparisons between render engines. I'm posting all these findings because I think it will improve Corona as a render engine. Corona Team, you know I love you.

fStorm split the HDRi into indirect and direct pass, and I think this is the reason fStorm has different shadow values. I've talked to the Corona dev team, and they have ensured me that Corona use the correct method. So ignore the value difference, this post is about the diffuse roughness. I've made my own custom Corona to fstorm and fstorm to Corona converter, and when it comes to interior renders, diffuse roughness has a lot to say on the final render. This is a parameter that should not be overlooked by the dev team.
             ___
    _] [__|OO|
   (____|___|     https://www.twitch.tv/dubca7 / https://soundcloud.com/dubca7 / https://dubcatshideout.com  ( ͡° ͜ʖ ͡°)

2018-01-26, 01:54:00
Reply #125

Dionysios.TS

  • Active Users
  • **
  • Posts: 524
    • View Profile
    • Personal Portfolio
+1
Responsable d'Imagerie
Renzo Piano Building Workshop / Paris

https://dionysios.myportfolio.com/

2018-01-26, 15:24:48
Reply #126

Fluss

  • Active Users
  • **
  • Posts: 408
    • View Profile
Dubcat I don't understand what you're trying to demonstrate in your last post. Can you explain a little bit further pls?

2018-01-26, 15:43:34
Reply #127

romullus

  • Global Moderator
  • Active Users
  • ****
  • Posts: 5724
  • Let's move this topic, shall we?
    • View Profile
    • My Models
Open images in new tabs and switch between them. Look at sphere's shadow terminator - the difference should be obvious.
I'm not Corona Team member. Everything i say, is my personal opinion only.

2018-01-27, 17:22:25
Reply #128

dubcat

  • Active Users
  • **
  • Posts: 456
  • ฅ^•ﻌ•^ฅ meow
    • View Profile
Can you explain a little bit further pls?

As romullus said :)

--------------------------------------------------------------

I'm posting a little weekend present for you guys, these settings emulate the default ACES tonemapper with about 0.5 RGB error.
The settings might look ridiculous with negative "highlight compression", but trust me, give it a try ;)



I'm attaching a VFB settings file that you can load.
Remember to disable LUT if you are using one.

edit: I made these settings with "Corona 2018-01-25 Daily". I remember there were some filmic changes, but I can't remember if it was 1.6/1.7 or daily. Just letting you know in case stuff looks weird. The result should be pretty much identical to the ACES Photoshop script I posted earlier.
« Last Edit: 2018-01-28, 03:59:30 by dubcat »
             ___
    _] [__|OO|
   (____|___|     https://www.twitch.tv/dubca7 / https://soundcloud.com/dubca7 / https://dubcatshideout.com  ( ͡° ͜ʖ ͡°)

2018-01-28, 15:42:46
Reply #129

agentdark45

  • Active Users
  • **
  • Posts: 462
    • View Profile
I just tried out the above settings (with tweaked exposure) on some previous CXR's and I have to say I'm super impressed with the result!

Straight out of the VFB I'm getting some very pleasing images. Nice natural burnouts, good contrast/shadows and rich colours. We are definitely on to something here.

Comparison images below:

1. Default Corona settings with highlight compression at 8
2. Manually tweaked settings (no lut)
3. Manually tweaked settings (with lut)
4. Dubcat's settings
« Last Edit: 2018-01-28, 15:56:19 by agentdark45 »
Vray who?

2018-01-28, 21:31:23
Reply #130

Fluss

  • Active Users
  • **
  • Posts: 408
    • View Profile
Open images in new tabs and switch between them. Look at sphere's shadow terminator - the difference should be obvious.

Indeed, the transition between light and shadow is way softer in Fstorm. Much more pleasing to the eye.

How does the Vray alSurface shader does compare to this?

"The Fresnel effect is computed as part of the BRDF calculations (a.k.a. “glossy Fresnel”) and takes into account the viewing direction, the surface normal, and the light directions."

Isn't it related to this ?
nothing to do with the diffuse part, misread what you wrote

Also, As far as I remember, Corona shader does not support glossy fresnel.




What's more, some users are experiencing strange atifacts with SSS here :

https://corona-renderer.com/forum/index.php?topic=18702.0

Which seems to be the exact same issue descibed here :

Quote
VRay Skin Mtl: This uses a two lobe specular model with a sharp and broad reflection. It also uses Phong as the reflective BRDF. This model proved to work well in many situations, but at glancing angles, based on the way that it would cut out the SSS, could cause darkening. This is because the Fresnel effect is computed as a function of the angle between the viewing direction and the surface normal, ignoring the directions from which lights illuminate the surfaces.

alSurface: This model also uses a two lobe specular model. However, instead of using a smooth BRDF like Phong or Blinn, it uses a microfacet one. The Fresnel effect is computed as part of the BRDF calculations (a.k.a. “glossy Fresnel”) and takes into account the viewing direction, the surface normal, and the light directions. The user has a choice between GGX and Beckmann BRDF models. Based on the nature of the micro faceting, it can avoid the darkening effect at the glancing angles through retro-reflection. Additionally it does not cover the SSS at the same glancing angles.

source : https://www.chaosgroup.com/blog/v-rays-implementation-of-the-anders-langlands-alsurface-shader

Of course, I might be totally misleading as i'm not a tech guru like some of you guys..

But there is one thing that bother me. We were experiencing some white halos on rough shaders some times ago. By looking at the examples here ( https://www.chaosgroup.com/blog/understanding-glossy-fresnel), it looks like a direct consequence of not supporting the glossy fresnel effect. It seems fixed now. But when i see that glossy fresnel isn't supported yet, i wonder if it has been fixed the correct way or with some workarounds (we have then seen some dark halos when the devs were trying to fix it, which strengthens me in my view that it wasn't done the right way).

Basically, initial implementation of the shader was good, it should be more reflective at grazing angles like it was doing BUT this effect should also be affected by the roughness as more micro-facets are facing the camera (so less grazing angles) on rough materials compared to shiny ones. If the whole thing is well implemented, angle affect should resolve by itself I guess.

fStorm split the HDRi into indirect and direct pass, and I think this is the reason fStorm has different shadow values.

That said, your statement totally make sense here.
« Last Edit: 2018-02-03, 16:48:51 by Fluss »

2018-01-29, 09:16:51
Reply #131

NicolasC

  • Primary Certified Instructor
  • Active Users
  • ***
  • Posts: 386
  • Primary Certified Instructor
    • View Profile
    • Intangibles Assets Design
Thanks A LOT Dubcat for all your investigations. Very instructive, a gold mine.
Hats off, really.
Nicolas Caplat

3d Dept Manager / Intangibles Assets Design / Paris

2018-01-29, 17:05:05
Reply #132

Juraj Talcik

  • Active Users
  • **
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
Quote
it has been fixed the correct way or with some workarounds

It was some sort of workaround by manually fitting it to resemble the alsurface curve. And the diffuse part is still basic Lambert, which is responsible for the harsch look.

Ondra said he would like to look into this more now, not sure when that is but we should keep the pressure and push him towards it :- ).
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-01-31, 00:53:07
Reply #133

Dippndots

  • Active Users
  • **
  • Posts: 289
  • Alex Fagan Co-Founder at The Faction
    • View Profile
    • The Faction
Is your Megascans export script still available, Dubcat? The link in your Megascans thread 404s :(

2018-01-31, 10:15:48
Reply #134

Fluss

  • Active Users
  • **
  • Posts: 408
    • View Profile
Quote
it has been fixed the correct way or with some workarounds

It was some sort of workaround by manually fitting it to resemble the alsurface curve. And the diffuse part is still basic Lambert, which is responsible for the harsch look.

Ondra said he would like to look into this more now, not sure when that is but we should keep the pressure and push him towards it :- ).

Hopefully, corona 2.0 will bring Vray compatibility and OSL support so we might be able to try that alshader inside corona sooner than later.