Author Topic: Time to ditch sRGB/Linear as default (?)  (Read 37147 times)

2018-10-19, 17:09:02
Reply #120

Jpjapers

  • Active Users
  • **
  • Posts: 1215
    • View Profile
This thread kinda burned out but it was really interesting to read. Did this ever get picked up as a potential feature request? It probably should.
when is the last time we had a poll about what users want?

Umm... always? There's always one - https://corona-renderer.com/forum/index.php?topic=96

Tom do you think reworking tonemapping could be added to the poll?
I see its on the future ideas board on the trello but it seems from this thread that theres a huge appetite for things to be changed on that front sooner rather than later.

2018-10-20, 03:39:32
Reply #121

Njen

  • Active Users
  • **
  • Posts: 501
    • View Profile
    • Kickstart Cyan Eyed!
If I look at the roadmap I see all kinds of improvements that are interesting but it's not the things uses want (I think),
Not what users what? What are you talking about?

Users want: caustics, clearcoat, fixed volume media, removing portals, cryptomatte, and more.

I think when you said "it's not the things uses want (I think)", you actually meant "it's not things I want".
Please support my Kickstarter for my animated film, Cyan Eyed (rendered in Corona)!

2018-10-20, 12:46:12
Reply #122

Tok_Tok

  • Active Users
  • **
  • Posts: 84
    • View Profile
If I look at the roadmap I see all kinds of improvements that are interesting but it's not the things uses want (I think),
Not what users what? What are you talking about?

Users want: caustics, clearcoat, fixed volume media, removing portals, cryptomatte, and more.

I think when you said "it's not the things uses want (I think)", you actually meant "it's not things I want".

I'll be blunt. The whole race, since the beginning of CGI, is to make the most realistic image possible. That's what you see people do and admire when somebody is making images that are next level in realism (and also in beauty, but that's dependent on the user not the engine). So the MOST important thing for a render engine is to keep up with what's technical possible. At the moment Fstorm is producing more realistic images than Corona, don't ask me why or how, I'm not a technician, it just does. So I think that the most important thing for the Corona team is to improve the engine to be on par with the best. I think THAT is the most important thing to work out now, and that's not in the poll.

The whole reason I switched, a few years ago from, Maxwell Render to Corona is that Corona is better (read more realistic and faster). If you ask me, Maxwell is nice but it's dying out because they did not improve their render model and kept to the physically correct rendering of light. I love corona(!) so I would hate to see the same thing happening to them so I really like to see that they keep improving their render quality so it still on par with the best, as they used to be a few years back.

I know a lot of people not going to agree with me because this might be too bluntly said, but I think that in it's core this is true. People will eventually chose for what's the best quality. :)

2018-10-20, 14:25:38
Reply #123

monitorhero

  • Active Users
  • **
  • Posts: 13
    • View Profile
If I look at the roadmap I see all kinds of improvements that are interesting but it's not the things uses want (I think),
Not what users what? What are you talking about?

Users want: caustics, clearcoat, fixed volume media, removing portals, cryptomatte, and more.

I think when you said "it's not the things uses want (I think)", you actually meant "it's not things I want".

I'll be blunt. The whole race, since the beginning of CGI, is to make the most realistic image possible. That's what you see people do and admire when somebody is making images that are next level in realism (and also in beauty, but that's dependent on the user not the engine). So the MOST important thing for a render engine is to keep up with what's technical possible. At the moment Fstorm is producing more realistic images than Corona, don't ask me why or how, I'm not a technician, it just does. So I think that the most important thing for the Corona team is to improve the engine to be on par with the best. I think THAT is the most important thing to work out now, and that's not in the poll.

The whole reason I switched, a few years ago from, Maxwell Render to Corona is that Corona is better (read more realistic and faster). If you ask me, Maxwell is nice but it's dying out because they did not improve their render model and kept to the physically correct rendering of light. I love corona(!) so I would hate to see the same thing happening to them so I really like to see that they keep improving their render quality so it still on par with the best, as they used to be a few years back.

I know a lot of people not going to agree with me because this might be too bluntly said, but I think that in it's core this is true. People will eventually chose for what's the best quality. :)

Right now I think Fstorm and Corona are on par. It might take a little more effort to achieve the same results in Corona. What I miss from Fstorm which I think is crucial is triplanar mapping. And it's not a planned feature in the near future. (I asked Andrey) Is there another way to achieve that in Fstorm without unwrapping?

But I still would prefer the ease of the fstorm tonemapping.

2018-10-20, 21:40:32
Reply #124

Njen

  • Active Users
  • **
  • Posts: 501
    • View Profile
    • Kickstart Cyan Eyed!
If I look at the roadmap I see all kinds of improvements that are interesting but it's not the things uses want (I think),
Not what users what? What are you talking about?

Users want: caustics, clearcoat, fixed volume media, removing portals, cryptomatte, and more.

I think when you said "it's not the things uses want (I think)", you actually meant "it's not things I want".

I'll be blunt...

I can be blunt too. Most of the things on the poll are to achieve realism, so I'm not too sure what your point is.
Please support my Kickstarter for my animated film, Cyan Eyed (rendered in Corona)!

2018-10-22, 10:51:28
Reply #125

Jpjapers

  • Active Users
  • **
  • Posts: 1215
    • View Profile
I know id much rather see an improvement to my workflow and realism right out of the VFB is all and it seems like theres pretty strong opinion from this now 9 page thread that this 'feature request' is something the user base really does want. But I agree with the above, Its not on the poll and it should be. Ive mentioned on the poll thread that stuff like 'speed improvements' and 'memory improvements' shouldnt be on the poll because version to version im pretty certain we all expect those things. Id rather see some new tonemapping than be able to render ever so slightly faster if im completely honest.
« Last Edit: 2018-10-22, 11:10:57 by jpjapers »

2018-10-22, 12:37:16
Reply #126

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 8899
  • Turning coffee to features since 2009
    • View Profile
I added the tone mapping into the list. It is high on my list of features to do, the problem is that we are talking about very subtle differences and as Tok_Tok said, we dont even know why some tone mappers look more realistic than others. What can be definitely improved though is usability - removing current controls for something like shadows/midtones/highlights controls
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2018-10-22, 13:00:11
Reply #127

Jpjapers

  • Active Users
  • **
  • Posts: 1215
    • View Profile
I added the tone mapping into the list. It is high on my list of features to do, the problem is that we are talking about very subtle differences and as Tok_Tok said, we dont even know why some tone mappers look more realistic than others. What can be definitely improved though is usability - removing current controls for something like shadows/midtones/highlights controls

Oh its definitely subjective for sure but thank you for letting us know. Im sure alot of people would be very pleased.
I wonder if theres a way of measuring the effects of various tonemappers. Im sure dubcat or someone could find a way haha

2018-10-22, 13:07:35
Reply #128

pokoy

  • Active Users
  • **
  • Posts: 1420
    • View Profile
Shadow/midtones/highlights controls would be fantastic!

2018-10-24, 08:41:33
Reply #129

monitorhero

  • Active Users
  • **
  • Posts: 13
    • View Profile
I added the tone mapping into the list. It is high on my list of features to do, the problem is that we are talking about very subtle differences and as Tok_Tok said, we dont even know why some tone mappers look more realistic than others. What can be definitely improved though is usability - removing current controls for something like shadows/midtones/highlights controls

Why remove those features? I think they are useful. But I am glad to hear that it's high up on your list. And definitely talk to Dubcat about that. I think he can offer a lot of insight into this :-) please :D

2018-10-24, 10:32:00
Reply #130

romullus

  • Global Moderator
  • Active Users
  • ****
  • Posts: 5893
  • Let's move this topic, shall we?
    • View Profile
    • My Models
They would be more usefull if they would work well. I often find filmic shadows and filmic highlights to work in weird, unpredictable fashion. Solid highlight/midtones/shadows controls are much needed in Corona. Can't wait for them.
I'm not Corona Team member. Everything i say, is my personal opinion only.

2018-10-24, 12:10:07
Reply #131

kosso_olli

  • Active Users
  • **
  • Posts: 31
    • View Profile
...we dont even know why some tone mappers look more realistic than others.

A user here on the forums (dubcat I believe) posted a way to analyze two images. The result was a curve that represents the difference between the two. So if you have a raw linear render from fStorm for example, and another tonemapped one, you could try to analyze what is going on there by looking at the output curve. Just an idea, maybe ask dubcat... I am highly interested in this as well, because I feel the 32bit linear workflow is not suited for the needs of most users here.
The V-Ray guy checking out Corona...

https://www.behance.net/Oliver_Kossatz

2018-10-24, 15:14:50
Reply #132

Fluss

  • Active Users
  • **
  • Posts: 423
    • View Profile
...we dont even know why some tone mappers look more realistic than others.

A user here on the forums (dubcat I believe) posted a way to analyze two images. The result was a curve that represents the difference between the two. So if you have a raw linear render from fStorm for example, and another tonemapped one, you could try to analyze what is going on there by looking at the output curve. Just an idea, maybe ask dubcat... I am highly interested in this as well, because I feel the 32bit linear workflow is not suited for the needs of most users here.

There is no need to analyze anything here we know what happen on the Fstorm side. It's using the ACES output transform instead of the sRGB one. Tonemapping controls can be whatever you want before applying the output transform (which must be the final step), it's just a matter of taste.

2018-10-25, 20:21:45
Reply #133

monitorhero

  • Active Users
  • **
  • Posts: 13
    • View Profile
A thing I just read that I find very interesting and also raises a few questions about ACES and RGB in general:

http://nbviewer.jupyter.org/gist/sagland/3c791e79353673fd24fa

https://www.colour-science.org/posts/about-rendering-engines-colourspaces-agnosticism/

I guess spectral rendering is the future otherwise we will always rely on cheats/artistry ;)

2018-10-25, 21:42:54
Reply #134

Juraj Talcik

  • Active Users
  • **
  • Posts: 3656
  • Tinkering away
    • View Profile
    • studio website
The Filmic HL/Shadows (in their current implementation) should be removed. I understand some people use them to get interesting effect when it occassionally does what they 'think' it should do, but it never actually does that. The implementation is wrong so I would not hesitate to remove it fully.


https://www.colour-science.org/posts/about-rendering-engines-colourspaces-agnosticism/


This one is actually very interesting ! So wide gamut color space as internal computation space can actually provide problematic output when used with common lighting engine ? Not sure how this transfers to Corona but it's interesting as I never thought this could affect that much.
« Last Edit: 2018-10-25, 21:46:41 by Juraj Talcik »
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika