Author Topic: Here's a hard one. Volumetric lightning / GI FOG  (Read 7669 times)

2014-03-14, 19:53:33

unknownhumanoid

  • Users
  • *
  • Posts: 2
    • View Profile
First of all I'm new here and new to Corona also. It's a very interesting rendering engine and will probably use it in the future for production purposes.

The story of the volumetric lightning, or GI fog, how do you want to call it:
This "little" feature was introduced in a pre-alpha version of Maxwell Render few years ago. Since was demanding large ammount of processing (some little tests with poor quality took even 200h) they took it out and not even today we don't have such a thing.

So it's not about simple fake volumetric light not even close with the fake fog simulated with zdepth channels in PH.

I'm talking about real fog. The one which creates the so called "volumetric light" when the rays spread and illuminate the particles inside the fog and disperse the light in all directions.
So in a real environment a dusty air disperse the light inside the volume. A certain number of rays must be calculated creating particles in the atmosphere.

I'm still waiting, maybe someday, someone, if we are ready, will create a realistic way to simulate this kind of thing. I guess this is the top notch feature and the last one in the whole industry.

PS: if someone tries to convince me it was already made, believe me. It's all fake simulation. NextLimit tried and it was a fail; we dont have computers for such a thing. Not yet, or maybe there could be a different way.
 A real VL would expand your rendering time 100x times. Probably. Think about: every soft/reflection will disperse the light in directions which would illuminate particles which disperse the light further and reflect it also back onto surfaces, the fog itself becoming an illuminating source.
In fact the "Volumetric Light" doesn't exist. It's an effect resulted from what I described above. More correct would be GI FOG for example.
« Last Edit: 2014-03-15, 12:08:22 by unknownhumanoid »

2014-03-14, 20:59:21
Reply #1

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 13655
  • Marcin
    • View Profile
A bit offtopic but Terragen has some really good volumetric effects but they are probably faked in some degree. You can download free version of Terragen from http://planetside.co.uk/ . Seeing Corona as a renderer in Terragen would be something extremely cool but I doubt it will ever happen.
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2014-03-14, 21:06:07
Reply #2

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
Unbiased light transport in participating media is a really hard problem, because you have to integrate (compute) the illumination not only on scene surfaces (finite number of points per each sample), but also along the entire lines in space. This adds an extra dimension to the problem, making the render times skyrocket.

There are several combinations of the problem (number of bounces, homogenous vs. heterogenous (density-mapped) media, isotropic vs. anisotropic scattering, etc.), some of them are not currently practically solvable while others are (maybe maxwell cannot do it, but there are research renderers that can). Additionally, volume rendering has its specific challenges, that were ignored by the research community until recently. In last few years, some papers popped about volumetric rendering, some of them being potential game changers - for example one from last year here: http://cgg.mff.cuni.cz/~jaroslav/papers/2013-jis/index.htm reports a 1444× speedup in some situations.

BTW: Jarda, co-author of that paper, is joining the Corona team.
Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)

2014-03-14, 21:31:23
Reply #3

unknownhumanoid

  • Users
  • *
  • Posts: 2
    • View Profile
Incredible. Maybe in the near future we will have a solution.
What about limiting this to a given volume resolution (density) and a given number of bounces? Maybe a solution made to scale down progressively the calculation would give a reliable solution for everybody.
Maybe, I'm just saying. I'm not a programmer.

2014-03-14, 23:34:35
Reply #4

Ludvik Koutny

  • VIP
  • Active Users
  • ***
  • Posts: 2557
  • Just another user
    • View Profile
    • My Portfolio
Incredible. Maybe in the near future we will have a solution.
What about limiting this to a given volume resolution (density) and a given number of bounces? Maybe a solution made to scale down progressively the calculation would give a reliable solution for everybody.
Maybe, I'm just saying. I'm not a programmer.

This technique is common for all production renderers out there, and already used in production. You can generate volume fog with real light shafts. No one uses just zdepth fake anymore. Vray has VrayFog, Mental Ray has parti_volume, Arnold has volume effects as well...  even new Maxwell 3 has some. It does real volume with scattering. The only difference from unbiased solution is that the volume does not have infinite resolution... but still, it is sufficient for most of the effects.

2014-03-15, 03:24:58
Reply #5

Juraj

  • Active Users
  • **
  • Posts: 4815
    • View Profile
    • studio website
Incredible. Maybe in the near future we will have a solution.
What about limiting this to a given volume resolution (density) and a given number of bounces? Maybe a solution made to scale down progressively the calculation would give a reliable solution for everybody.
Maybe, I'm just saying. I'm not a programmer.

This technique is common for all production renderers out there, and already used in production. You can generate volume fog with real light shafts. No one uses just zdepth fake anymore. Vray has VrayFog, Mental Ray has parti_volume, Arnold has volume effects as well...  even new Maxwell 3 has some. It does real volume with scattering. The only difference from unbiased solution is that the volume does not have infinite resolution... but still, it is sufficient for most of the effects.

I second this. VrayEnviroFog for example works quite excellently, far from "fake" (well, everything by Maxwell marketing team terminology is fake). Though if one wants to use it in full "raw" nature in default state (using atleast 100 GI bounces with GI ticked obviously on, enough samples to get very clean result, mapped with some procedural maps, etc... and baked directly into "beauty" scene of complicated space, the render time will also multiply between 10-100. So there's that, you still have to use its within its limitations and intended tricks.
Please follow my new Instagram for latest projects, tips&tricks, short video tutorials and free models
Behance  Probably best updated portfolio of my work
lysfaere.com Please check the new stuff!

2014-06-10, 12:30:28
Reply #6

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 9048
  • Turning coffee to features since 2009
    • View Profile
Rendering is magic.How to get minidumps for crashed/frozen 3ds Max | Sorry for short replies, brief responses = more time to develop Corona ;)