View topic - Test lighting

Test lighting

Only for showing your progresses and for art related questions. Not for asking about usage.

Re: Test lighting

Post Mon Jul 02, 2012 9:11 pm

Alain wrote:Why is it so bright behind Monalisa ?

Kind regards
Alain


oops! My fault!. :oops: For some reason I removed some geometry from that wall before saving the blend file and the render was using only a plane as a wall (screwed up geometry :P). That make the light leaks from the outside through the Wall (just a plane).
I rendered the scene again this time with proper geometry for Walls, floor and ceiling. ;)
Attachments
interior_ver2_final.png
EXR file tonemapped with Luminance 2.3 Beta1, postproduction with GIMP
interior_ver2.jpg
default render from Yafaray in LDR format

"Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men."

User avatar
condar
 
Posts: 771
Joined: Sun Apr 12, 2009 2:41 am

Re: Test lighting

Post Mon Jul 02, 2012 9:23 pm

Abner, much better :)

Alain, you are absolutly right.

Condar, your advices are very important.
Your scene without previous issue is perfect.
I also agree with your vray IC observations.
Can you also share your post pro procedure in detail?

Once again, thank you all for sharing your work.
Arquitecto can be contacted here:
Arquitectos
Arquitecto
 
Posts: 249
Joined: Mon Jul 27, 2009 8:42 am
Location: Portugal

Re: Test lighting

Post Mon Jul 02, 2012 11:21 pm

Arquitecto wrote:Can you also share your post pro procedure in detail?

Once again, thank you all for sharing your work.


1.- I usually choose tonemap in Luminance with all the operators available from the dropdwon list, using the inverse value of Gamma Correction in Yafaray as PreGamma in Luminance Interface.
2.- Then I just mix all or some of the tonemapped images over the original render using blend modes in GIMP (screen, Overlay, etc) and using layermasks if it's needed.
3.- AO pass (multiply) + Zdepth pass (inverse Screen)
4.- Add Background if it's needed
5.- Color Correction + some filter you like

BUT it's always a matter of personal taste.
Attachments
Luminance.png
Luminance Interface with PreGamma = 0.45
Tonemap_Operators_result.jpg
Tonemapped images from Luminance
GIMP.png
composing all of some of the images in GIMP

"Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men."

User avatar
condar
 
Posts: 771
Joined: Sun Apr 12, 2009 2:41 am

Re: Test lighting

Post Tue Jul 03, 2012 10:12 am

:D
Thank you very much Condar.
Interesting aproach, in fact with lots of tonemapped images you don't loose information and you have light properties allways available on the gimp image.
An we definetly see that on the final image.
I still have one doubt.
You didn't make any AO pass and z-depth on the final image right? You were telling just your usual workflow right?
I am saying this because on production environment we need lens and some defocus, but doing it on yafaray or even vray is a never ending story to clean the noise... So I used to do it on photoshop. The last time I tried on gimp, i made a mask and then I used gaussian blur over the mask. But I wasn't quite happy with the results...Is this your usual aproach?

P.S.:Thanks again for sharing your knowledge. Suomi once explained his post pro procedures and I became quite surprised by his fantastic aproach. Rio had also a very nice different aproach. Post pro is a must on any good rendering and should be litle different from each render engine you use. For me it's not only taste... it's wisdom.
Arquitecto can be contacted here:
Arquitectos
Arquitecto
 
Posts: 249
Joined: Mon Jul 27, 2009 8:42 am
Location: Portugal

Re: Test lighting

Post Tue Jul 03, 2012 7:13 pm

Arquitecto wrote:I still have one doubt.
You didn't make any AO pass and z-depth on the final image right?


Yes I did "3.- AO pass (multiply) + Zdepth pass (inverse Screen)" but not always needed. As a general rule try to get all render passes you can and then decide which one you can use. In this case Zdepth inverted and blended with Screen 10% just to give the image a little sense of Mist. But Zdepth can be used to add Depth of Field too. To get AO pass and Zdepth in Yafaray is very simple and it only takes a few seconds to render: just remove all lights from the scene, render with Direct Lighting, activate AO and Clay render (pure white color), change the World Color to Pure White and render. Zdepth pass is only available in Blender exporter IF you change the output method from "Into Blender" to "Image File".

Arquitecto wrote:I am saying this because on production environment we need lens and some defocus, but doing it on yafaray or even vray is a never ending story to clean the noise... So I used to do it on photoshop. The last time I tried on gimp, i made a mask and then I used gaussian blur over the mask. But I wasn't quite happy with the results...Is this your usual aproach?


nope. Gimp have the "Focus Blur" Filter that can use a Zdepth pass as source to blur the main image. You just load a Zdepth pass in a different layer and then use it in the filter, select the "focus distance" and all the image will be blured proportionaly to to the Zdepth information and the max Blur radius. You can do something similar with Blender Nodes but in GIMP the process is more flexible and quite easy. In Blender, the focus distance is tighted to the camera information and so you can not change it as you want with nodes.
As an example I added DOF to the final image with an exagerated Blur Radius (max Blur 40) an a focus distance on the Red Chair just to demonstrate how the filter works. It looks terrible :D But you can note how the blur is distributed over the image according to the Zdepth information and the focus distance. In this scene considering the camera focus, a blur radius 1 or 2 will be enough to get a nice DOF effect.

Arquitecto wrote:Suomi once explained his post pro procedures and I became quite surprised by his fantastic aproach. Rio had also a very nice different aproach.


if you think about it, almost all post production processes are the same in a general way. the main differences are basically in the software people use wich determine their workflow. As a Linux user, I only use Open Source Software and GIMP cannot handle HDR image formats...for now (16 bit color depth images support in GIMP will be added in version 3.0) so I use Luminance or pfstool/pfstmo for tonemapping as an intermidiate step before blend and color correct the whole image. On the other hand people using Phtoshop can use the EXR file directly (I'm not sure that the same tonemapping filters are available in Photoshop) and so their workflow will be different without the intermediate step of getting tonemapped images but tonemapping in the final step inside Photoshop.
Attachments
AOandZdepth.png
AO pass and Z-depth pass from yafaray
Focus_Blur_Filter.jpg
Focus blur filter in GIMP
DOF_example.jpg
exagerated foucs blur example

"Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men."

User avatar
condar
 
Posts: 771
Joined: Sun Apr 12, 2009 2:41 am

Re: Test lighting

Post Wed Jul 04, 2012 10:14 am

:D
Thank you very much Condar!
This tutorial is so cool that could be sent to the tutorials section!
I Didn't knew I could make ao and z-deph pass with yafa.
I missed the output method change...

Focus blur is in fact a very nice filter.
Much better than my awfull aproach!
Your result is perfect! For testing purposes, of course! :D It looks like I am a giant looking inside a model :lol:
I also don't see any problem in missing 16 bit color deph on gimp post pro.
It seems gimp has much more potential!
And I must study it better... :wink:

Anyway, sincere thanks Abner for sharing your work, Allain for the critics, and Condar for the tuts!

P.S.: You guys do know the Mona Lisa real size, don't you? :lol:
Arquitecto can be contacted here:
Arquitectos
Arquitecto
 
Posts: 249
Joined: Mon Jul 27, 2009 8:42 am
Location: Portugal

Re: Test lighting

Post Wed Jul 04, 2012 10:59 am

@Condar
Very nice Method ! I didn't know about the Focus Blur Filter in Gimp :-)

I did a little test with Cycles (Postproductions with the Blender Compositor).
I know it's still noisy (1000 Samples, GPU, 13 min. Rendertime on a Quadro 5000).

render_01_040712.JPG


Kind regards
Alain
Intel Pentium 2.8 GHz 2 Cores, 8 GB RAM, GeForce GTX 480, Blender 2.63a, Win 7 64 Bit
Alain
 
Posts: 139
Joined: Thu Oct 23, 2008 3:34 pm

Re: Test lighting

Post Wed Jul 04, 2012 4:27 pm

Alain wrote:I did a little test with Cycles (Postproductions with the Blender Compositor).
I know it's still noisy (1000 Samples, GPU, 13 min. Rendertime on a Quadro 5000).

Wow, 13 mins on 350 cores and still not clean? So how long does it take on cpu? Surely it is less than 20 hours ?
User avatar
stoneage
 
Posts: 1241
Joined: Sun Dec 23, 2007 1:06 pm

Re: Test lighting

Post Thu Jul 05, 2012 7:33 pm

8 mins. on my GeForce GTX 480 :-)

Kind regards
Alain
Intel Pentium 2.8 GHz 2 Cores, 8 GB RAM, GeForce GTX 480, Blender 2.63a, Win 7 64 Bit
Alain
 
Posts: 139
Joined: Thu Oct 23, 2008 3:34 pm

Re: Test lighting

Post Thu Jul 05, 2012 9:02 pm

Alain wrote:8 mins. on my GeForce GTX 480

That is great news for you, but for the people who don't have upmarket GPUs to play with - are they looking at render times of ~ 10 - 20 hours per frame using Cycles?
User avatar
stoneage
 
Posts: 1241
Joined: Sun Dec 23, 2007 1:06 pm

Re: Test lighting

Post Fri Jul 06, 2012 12:38 am

stoneage wrote:
Alain wrote:8 mins. on my GeForce GTX 480

That is great news for you, but for the people who don't have upmarket GPUs to play with - are they looking at render times of ~ 10 - 20 hours per frame using Cycles?


I think you have to have a pretty crappy computer to get 10-20 hours/frame and if it's that bad i think you'll get insane render times with Yafaray as well.

I'm on an old computer from 2006 (2x2 core 2,66 Xeon) and i get quite long render times what ever renderer i use.
I love Yafaray but get frustrated working in blind, i change settings and have to re render to see the changes.
Since i don't have a cuda card i can't use Cycles on GPU either, but i can see all changes in real time which is a real time saver.
All scenes are different ant takes very different amount of time to render and get nice and clean.

But for this scene i was surprised how much time i had to to little tweaks and render tests to get a nice result with Yafaray and it ended up between 80-130 min to get it clean and noise free with PM.
So i tested it with Cycles for fun and got a clean render on my old CPU after 70-90 min.
I confess i am quite new to yafa but spent two days testing this scene over and over to tweak everything and rebuilt all the weird object that behaved strange and let light in behind the painting and in the book shelves.

These images should not be compared to each other since i changed everything between the scenes, the Yafa version use a blurred city background which i forgot to do for the Cyles version but both is straight out of Blender, only post work is the node editor.
Nice to have a AO and shadow pass node in Cycles these days.
This was just a test of what i could squeeze out of my old comp and try to get clean Yafa renders at minimum of time.

Thanks for sharing a cool scene and all the great tips shared in this thread.

Cheers

Here's the Yafaray 130 min
Yafa-130min.png


Here's the Cycles version at 90 min (nice reflections in the window)
Cycles-90min.png
Jakerlund
 
Posts: 77
Joined: Sun Aug 08, 2010 7:41 pm

Re: Test lighting

Post Fri Jul 06, 2012 2:40 am

Alain wrote:8 mins. on my GeForce GTX 480 :-)

Kind regards
Alain


7 min with my Asus GTX 580 but noisy as hell after 1000 samples per pixel ;)

"Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men."

User avatar
condar
 
Posts: 771
Joined: Sun Apr 12, 2009 2:41 am

Re: Test lighting

Post Fri Jul 06, 2012 3:12 am

But for this scene i was surprised how much time i had to to little tweaks and render tests to get a nice result with Yafaray and it ended up between 80-130 min to get it clean and noise free with PM.
So i tested it with Cycles for fun and got a clean render on my old CPU after 70-90 min.

Thank you, that is useful information.
The poor estimate is based on Alain's quote of 13 mins using 350 render cores to make a noisy image. Since no one was posting CPU render times, it was impossible to know whether this is scalable per render core.

The times for 1000 samples seem unrealistic since the render is not finished. I'm sure someone could make a noisy image with Luxrender or iRay in 5 minutes, but what would be the point? It really doesn't give any useful information.

Maybe a more useful comparison would be to include how many samples are required to produce a clean render.
User avatar
stoneage
 
Posts: 1241
Joined: Sun Dec 23, 2007 1:06 pm

Re: Test lighting

Post Fri Jul 06, 2012 8:31 am

stoneage wrote:
...
Maybe a more useful comparison would be to include how many samples are required to produce a clean render.


I think the more close comparison could be

Render Samples (Cycles) vs Path Samples*(Total Samples)(Yafaray)

with Total AA samples = (AA samples + Aditional Samples * (Passes - 1))

"Power tends to corrupt, and absolute power corrupts absolutely. Great men are almost always bad men."

User avatar
condar
 
Posts: 771
Joined: Sun Apr 12, 2009 2:41 am

Re: Test lighting

Post Fri Jul 06, 2012 9:53 am

condar wrote:I think the more close comparison could be

Render Samples (Cycles) vs Path Samples*(Total Samples)(Yafaray)

with Total AA samples = (AA samples + Aditional Samples * (Passes - 1))

Yes. 8)
User avatar
stoneage
 
Posts: 1241
Joined: Sun Dec 23, 2007 1:06 pm

Next

Return to Works in Progress



Who is online

Users browsing this forum: No registered users and 3 guests

cron