Author Topic: Juraj's WIPS/Theory/Questions //Update with few PDFs from my recent presentation  (Read 139728 times)

2013-09-25, 15:50:32
Reply #30

michaltimko

  • Active Users
  • **
  • Posts: 248
  • Coronaut (c)
    • View Profile
    • Michal Timko | Photorealistic architectural visualizations
Looks nice! Those pebbles / stones looks like creves from mastodont.
I think you used arroway, right? Something gray or blue-ish should work better imo.

That spotlight on upper left corner is intersecting with wall ?
Coronaut!(c)2011

www.michaltimko.net - Coming soon
Supporting Corona in commercial projects since pre-alpha Behance : www.behance.net/michaltimko/

2013-09-25, 16:44:29
Reply #31

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Looks nice! Those pebbles / stones looks like creves from mastodont.
I think you used arroway, right? Something gray or blue-ish should work better imo.

That spotlight on upper left corner is intersecting with wall ?

Heh, that's valid observation :- ) Funny thing is, those are already scaled down by 30perc. from the original size by Arroway. It's btw absolutely crazy package, there are like 100+ options...
I wanted to originally wanted to go with something white, than found it boring, but looks like you're right, because we also already observed the red color cast might be too much. Maybe the size too !

The spotlight isn't intersecting, it is such a model, partly inside ceiling, partly outside. This one : http://3dsky.org/modules/files/showfile.php?lid=50447
Just looks wrong maybe from this angle.
talcikdemovicova.com Website and blog

2013-09-25, 17:32:49
Reply #32

michaltimko

  • Active Users
  • **
  • Posts: 248
  • Coronaut (c)
    • View Profile
    • Michal Timko | Photorealistic architectural visualizations
Yea, that package is great. I have done some tests already.

Thanks for link, amazing model.
Coronaut!(c)2011

www.michaltimko.net - Coming soon
Supporting Corona in commercial projects since pre-alpha Behance : www.behance.net/michaltimko/

2013-09-25, 17:34:27
Reply #33

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Please publish the rendering time of your tests with MSI! Because increase MSI give more GI noise and need more rendertime!
And highlight strength can be increase by IOR or level/color reflection, but without raise the rendertime!
MSI is mostly a significant impact only on the caustic!

Incorrect, judge by yourself from the pictures. Each test run for same duration, 5 minutes. Low MSI cuts away lot of rays that would otherwise contribute significantly to GI and reflection.

Also, artificially rising reflection is WRONG solution, and it will not yield similar results, actually, far from it. Because if you raise reflectivity, but keep MSI low, the whole surface will reflect more light, e.g being lighter overall, the spectrum from dark to highlight will change. The proper gradient between shadows and highlights is achieved only with higher MSI.

But it doesn't matter to those who don't need it. It's there after all as "optimization" parameter, you trade of correctness for speed. If you want speed, you keep low MSI, if you want to get more correct, closer to unbiased look, you raise the value. But that is the only solution, not artificially creating materials with high albedo and unreal reflectivity.
talcikdemovicova.com Website and blog

2013-09-25, 17:39:45
Reply #34

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Yea, that package is great. I have done some tests already.

Thanks for link, amazing model.

It is great until you must choose, but this is first time when there are just too many options for sake of it :- ) It's absolutely crazy. And the size of it...hundreds of 4k textures.
Too bad the materials provided are for Maxwell. They have crazy nice demos rendered with it on website.
talcikdemovicova.com Website and blog

2013-09-27, 23:29:54
Reply #35

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Tried to optimize the scene as much as possible, to see how fast I can get away with. It's not overdone, as the glass (thick, hybrid mode) is still around room, though I could hide the one behind camera, I choose not to.
Interestingly, 95perc. of scene looks identical after 30 minutes and 3 hours alike for this 1920x1080px render. It's just the micro-noise in darker walls on left and right in the picture. It doesn't bother me much, but for commercial projects, I want to be as clean as possible. I guess here the adaptivity could somehow help, as most of the picture just get oversampled too much for nothing.
Anyway, for now, best result are using hi-res and downsampling with slight blur. I like the detail better than.

Btw settings: PT32 ( I tend to find the higher PTs to converge with much less fireflies or speckles at all, while noise seems to be identical, just less amount of passes, but since I never have problem with AA, neighter use DOF, this number seems fine for most of my scenes), Light Samples = 2 ( not sure if I should ever rise this for my work, from test, haven't really found better noise converge, I mostly need GI rather than direct light noise to clear out), HD Cache 2048, rest default. MSI 200

Not much time to test this, but does anyone use different filter outside of default Tent ? Any reason why and what are the results ?

talcikdemovicova.com Website and blog

2013-09-27, 23:35:09
Reply #36

Ondra

  • Administrator
  • Active Users
  • *****
  • Posts: 7501
  • Turning coffee to features since 2009
    • View Profile

Not much time to test this, but does anyone use different filter outside of default Tent ? Any reason why and what are the results ?

I use None, it is easier to debug scenes with :D
Rendering is magic.
Private scene uploader | How to get minidumps for crashed/frozen 3ds Max | Image saving/loading is handled by 3dsmax, not Corona

2013-09-27, 23:39:31
Reply #37

michaltimko

  • Active Users
  • **
  • Posts: 248
  • Coronaut (c)
    • View Profile
    • Michal Timko | Photorealistic architectural visualizations
Can you try bucket mode with 16 init samples and 2-4 passes (to fit your 30min time). ? Bucket have some adaptivity inside so...

Anyway, i have same problem with that noise which take ages to converge.
Coronaut!(c)2011

www.michaltimko.net - Coming soon
Supporting Corona in commercial projects since pre-alpha Behance : www.behance.net/michaltimko/

2013-09-27, 23:49:08
Reply #38

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Can you try bucket mode with 16 init samples and 2-4 passes (to fit your 30min time). ? Bucket have some adaptivity inside so...

Anyway, i have same problem with that noise which take ages to converge.

Should I do this within my current Alpha5 ? I will install daily in around 2 weeks, no courage right now :- ).
talcikdemovicova.com Website and blog

2013-09-27, 23:50:31
Reply #39

michaltimko

  • Active Users
  • **
  • Posts: 248
  • Coronaut (c)
    • View Profile
    • Michal Timko | Photorealistic architectural visualizations
Not sure but i think A5 is missing adaptivity ? Just open bucket tab and you`ll see :)
Coronaut!(c)2011

www.michaltimko.net - Coming soon
Supporting Corona in commercial projects since pre-alpha Behance : www.behance.net/michaltimko/

2013-09-27, 23:55:21
Reply #40

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website
Not sure but i think A5 is missing adaptivity ? Just open bucket tab and you`ll see :)

No it isn't there yet, I thought that you had in mind something about bucket mode directly. I had already played with adaptivity with latest daily, but somehow just wasn't sure it really did something, just seemed to only go slower. But I really didn't pay much attention to testing it, just can't afford the time now, but I was very curious. I replaced it again with Alpha5 to finish current projects.

I am somehow not really fit for tester, maybe I will just calmly wait for Alpha6. Time flies fast and I am quite happy with Alpha5.
talcikdemovicova.com Website and blog

2013-09-28, 00:55:17
Reply #41

Rawalanche

  • Global Moderator
  • Active Users
  • ****
  • Posts: 2448
  • Just another user
    • View Profile
    • My Portfolio
It is either noise from indirect illumination, or direct one (enviro light), or refractive or reflective caustics.

First one, indirect illumination kind, should get better with increasing amount of PT samples (Usually to something between 32 and 64).

Second kind...  direct illumination, will get better if light samples multiplier is increased (usually to something like 8).

Third, refractive caustics, can be minimized by either using hybrid glass mode and/or lowering MSI value.

Fourth kind, reflective caustics, can be again minimized by lowering MSI.

I would bet in this case that it is first or second. I recommend this:

Make a small region (150*100px at most) in the problematic spot

Set time limit to something like 20 seconds

Render and save the picture.

Increase PT samples, render again, save the picture, and compare. If this makes some noticeable difference, then it could be indirect illumination noise.

Revert PT samples back, increase light samples multiplier, render again, and compare with original. If this makes some noticeable difference, then it could be direct illumination noise.

Revert light samples multiplier back, select all glass object in scenes and hide them. Render again, and compare with original. This time to see if the problem is not refractive caustics. If this step has a huge impact, then it is probably glass.

If the noise is still there, and hiding glass did not make any significant difference, then unhide glass, lower MSI to something like 10, and render again. If that makes significant difference, then it is unclamped reflection caustics.

If none of these makes significant difference, then its just probably not that easy scene to render.

Other than that, it could be for example noise from bump map with way too low fitering. Or there can be a very little hole in the back room somewhere, where little bit of direct lighting leaks through from the environment and causes some noise... or there is always small possibility of it being a bug ;)
« Last Edit: 2013-09-28, 00:59:12 by Rawalanche »
Proud member of Cheeki Breeki brotherhood
https://www.artstation.com/artist/rawalanche

2013-09-28, 01:00:12
Reply #42

michaltimko

  • Active Users
  • **
  • Posts: 248
  • Coronaut (c)
    • View Profile
    • Michal Timko | Photorealistic architectural visualizations
Coronaut!(c)2011

www.michaltimko.net - Coming soon
Supporting Corona in commercial projects since pre-alpha Behance : www.behance.net/michaltimko/

2013-09-28, 01:20:35
Reply #43

rafpug

  • Active Users
  • **
  • Posts: 721
    • View Profile
It is either noise from indirect illumination, or direct one (enviro light), or refractive or reflective caustics.

First one, indirect illumination kind, should get better with increasing amount of PT samples (Usually to something between 32 and 64).

Second kind...  direct illumination, will get better if light samples multiplier is increased (usually to something like 8).

Third, refractive caustics, can be minimized by either using hybrid glass mode and/or lowering MSI value.

Fourth kind, reflective caustics, can be again minimized by lowering MSI.

I would bet in this case that it is first or second. I recommend this:

Make a small region (150*100px at most) in the problematic spot

Set time limit to something like 20 seconds

Render and save the picture.

........

...


holy words

2013-09-28, 01:23:11
Reply #44

Juraj_Talcik

  • Active Users
  • **
  • Posts: 2539
    • View Profile
    • studio website

Lots of nice advice

Hey, feeling alive at this time :- ) ? Thanks, thats sums it up nicely though ! It stays there regardless of MSI, and I only use Hybrid mode (even in scenes where the glass is just everywhere, even regular glass seems to do just fine, but I want to err on the safer side).
Now the playing between PT and SMs kinda annoys me, because I can't really decide when the settings are on the best spot and when it's overkill/under no matter how many tests I do. I can play forever with it :- D

But since you mention value 8 for SMs, that's not so high, and maybe I could just set this value there as default ? What is the range of values here that people use ? For example I will also include artificial lights here, where would be some sweat-spot to keep safe ?

Now, when I think of it, this noise appears only once I include CoronaSun as additional light to HDRi. Could this mean something ?
talcikdemovicova.com Website and blog