VRay-like interior renders with mental ray???

[dagon1978]

great works here

i hope everyone want to share results and settings!

keep it up!

ciau

mat

[Dankish]

Thanks guys.
And sure I’ll share the settings. At least that’s the way I learned a lot!

I’ve got 3 lights generally. The sun (directional) with no decay. Point light converted to mental ray area light at the window. 3000 photons, 1.9 exponent.
Two of them have the mib_cie_d mentalray shader. The third light is slightly behind the camera, set to 0 and photons settings I left as they are, default. Third light placed there because of the lack of details in some dark areas.)
Final Gather rays: 50; min radius: 60, max: 600. Secondary Diffuse Bounces turned on (I don’t know if using it along with GI has no effect, as some people say but I believe them I just haven’t tried rendering without it yet

About BSP. I am sure it’s already been commented, but here’s what I’ve done.
After the unsuccsessful render I did (in a manner of render speed) I took the output window to see the so called BSP Section. The values I used are the average depth and average leaf size.
And I think that’s pretty much it :).

RCI 0.2 info : main bsp tree statistics:
RCI 0.2 info : max depth : 49
RCI 0.2 info : max leaf size : 128
RCI 0.2 info : average depth : 28
RCI 0.2 info : average leaf size : 9
RCI 0.2 info : leafnodes : 32140
RCI 0.2 info : bsp size (Kb) : 1605

These values are just for example. They are different than the ones I used for my scene. As far as I know it’s a memory set up so It depends on different computers.
I don’t pretend to be good, so I might be wrong.

Here’s the lights concept

By the way: sorry for my english

[ronaldomiranda]

Viva La Mental Ray!!!

[Dankish]

And here’s the higher resolution image

[Splin]

You guys make me really longing after MR 3.4. Really good stuff here! Cant wait until 3.4 arrives to XSI.
Dankish, really great stuff. What I find little bit distracting is the reflection on the mirror. Left side of room seems to be not so bright but reflection shows that it is supposed to be really bright there. Same with bed etc, gamma seems to be bit off.
Cool stuff, geep rocking guys!

EDIT: Some time ago, Paul Debevecs light probe’s link was floating around in this thread. It was possible to download light samples from there. Over 60Mb or something. Really neat stuff and great idea but once I opened the package I didnt know anymore what to do. Those hdr’s are not full spheres but half spheres, how should one use them?
thx!

[Dankish]

Splin, do you mean this link http://www.debevec.org/Probes/
I think the IBL node wasn’t available in the versions previous 3.4 and I remember we had to create half a sphere and then put the HDRI into the ambient/incandescence channel. But now Image Based Lighting is featured in mental ray so there’s no problem.

And thanks for the kind words. I need to work a little more on the image :).

[Splin]

Ahh dang, sorry for confusion! Now I realized that this link was not posted in this thread. No I am not talking about the http://www.debevec.org/Probes/

If I remember it correctly I found it in Italian forum where dagon linked his interior tutorial.
Anyway, the link about I am talking here is about Debevecs last public project where he captured sky and sun from morning till to the sunset. Basically you get nice hdri sky maps in different hours. I downloaded the whole package but later found out that those are only half spheres. If someone knows what about I am talking here, that would be great! If not…well yeah forget it

[T-R]

It’s right here
http://www.ict.usc.edu/graphics/skyprobes/

The file is called skyprobes.zip at the bottom of the page.

[Splin]

It’s right here
http://www.ict.usc.edu/graphics/skyprobes/

The file is called skyprobes.zip at the bottom of the page.

Thx dude for the link! Yeah thats what I meant!
Since I can’t speak Italian I can not ask this question in that forum where Dagon posted his tutorial.
Anyway, one who downloads that package will find half cut hemispheres from there, not full hemispheres. It is interesting how one goes with applying them to the exterior scene. Is this some new technique or Debevec just didnt want to share the full thing. Still for lighting interior shots it could be useful…

[floze]

Is this some new technique or Debevec just didnt want to share the full thing.

You’d better take at look at the pdf files on the page T_R posted:

http://www.ict.usc.edu/graphics/skyprobes/skycapture_afrigraph2004.pdf

http://www.ict.usc.edu/graphics/skyprobes/ms_thesis_stumpfel.pdf

http://www.ict.usc.edu/graphics/skyprobes/skycapture_poster.pdf

and a neat little movie:
http://www.ict.usc.edu/graphics/skyprobes/skycapture_poster.mov

Paul Debevec does want to share the full thing. The point is that the skyprobes concentrate on capturing the full dynamic range of sun and sky.

[Splin]

Hey Floze…thx alot for reply!
Those pdf documents are very useful…to be honest I ran those thru already before. In all of those documents there are screenshots from the full hemisphere of sky capture shots. But still in sample package comes only half hemisphere version. Is MR unable to use those half hemisphere shots? I have some problems to understand the logic behind it. How do you apply only 180 degree hdri map to a scene what is actually 360 degrees. Some area will be black and FG will count that in, thus guiding to non correct renderings.
Sry if I am talking pure insanity here…if so just scream out loud RTFM
Have you used yourself yet those sky samples to light you renderings?

[Synthesizer]

I think they are only half because of the way they were shot. In that movie they have, it says “Fisheye sequence” my guess is that these were shot with a 180 degree fisheye, which can only get half, or they used that half ball on the black table thing. All you have to do is make a ground that will cover the bottom half up.
BTW, they are really bright! I had to tone them down with hdr shop to be able to use them without setting the colour gain to .00001 and the like.

[floze]

Sry if I am talking pure insanity here…if so just scream out loud RTFM
Have you used yourself yet those sky samples to light you renderings?

Ok, it’s time for some conceptual renderings again. Below are the results of a (more or less) quick test I’ve done today - sunday is testing day!
The scene is lit by an photon emitting IBL containing the skyprobe HDR images and using an interval of approximately 1h. The sunlight on the other hand is a directional light generated by the LightGen plugin (http://www.ict.usc.edu/~jcohen/lightgen/lightgen.html) for hdrshop (http://www.ict.usc.edu/graphics/HDRShop/). To ‘extract’ the sun out of the HDRI, I made a threshold of 1000 pixel values, converted the format from light probe to latitude/longitude (LightGen requires this) and configured the plugin to generate only 1 lightsource.
As you can see the detail of global illumination is not superior, but this is due to the fact I used only 25000 photons - raise this value sufficiently, and the detail will be satisfying. Also, the finalgathering could be optimized, but what the heck:

10-10

11-30

12-40

13-40

14-40

15-50

16-40

17-20

You may download the scene over here (1.85 MB): http://www.floze.de/dump/tut_jap_ibl.zip …I hope dagon doesnt mind…?

It’s absolutely not perfect, but I think you get the idea behind it.
The lack of a southern hemisphere is only noticeable in the lower part of the left window. If I did some serious renderings, I’d compose the background anyway.

Have a nice sunday! I hope your weather is as good as over here right now.

[floze]

Sorry for replying to myself, but I technically cant attach more images to the last post. Here’s a refined render of the 16-40 probe, using a hint of ambient occlusion and a bit depth of field:

I hope you dont mind the lack of textures, but my main intention is to show off the bare lighting with a least amount of distortions.

[Synthesizer]

Looking good Floze:thumbsup: Just wondering, you have the final gather colour gain set to 0, how come? Would this be the same as turing it off, or does it still do something? It’d be cool to see an animation of this, maybe I’ll try setting one up an letting it render overnight.

[dagon1978]

great work floze!
what about render times?
i like this approach

oh, a thanx for sharing!

[floze]

Looking good Floze:thumbsup: Just wondering, you have the final gather colour gain set to 0, how come? Would this be the same as turing it off, or does it still do something? It’d be cool to see an animation of this, maybe I’ll try setting one up an letting it render overnight.

Thx Synthesizer and dagon!
Good question Synthesizer. Finalgather color gain of the IBL node is set to black, because:

The captured lighting presented here could be used to render any outdoor scene, providing realistic illumination and a sky backdrop. However, due to the extreme contrast between the sun and sky, standard monte-carlo illumination algorithms will produce a large amount of noise, as they will have difficulty sampling the small disc of the sun reliably and sufficiently.[…]

If I used the IBL to contribute to the finalgathering process, I’d have (among others) another source of noise. So I simply switched the IBL’s fg contribution off. This does not affect the rest of the scene, since I’m using photontracing as source of indirect illumination, and finalgathering smoothens it out.

But: I could use the IBL to contribute to fg, I only needed to set a fg filter of at least 1. But I tend to avoid this filter strictly because eliminating the real source of error is imho the better choice. Another solution would be to take the HDRI, make a threshold of something between 1000 and 2000 pixel values, and subtract this result from the original image via the ‘Calculate’ operation in hdrshop. This way you could ‘subtract’ the sun from your HDRI, resulting in a similar fg calculation like with filter set to 1 or greater. I guess Paul Debevec used a similar method in his examples, dont know exactly though.
Funny enough this technique was intended to be used for outdoors only…

Well, with photons set to rebuild-off and a frozen or rebuild-off irradiance cache, rendering the scene is a piece of cake. Emitting the photons naturally takes some time because the whole IBL sphere needs to be sampled all over, only letting photons into the special cage I built. But the rest is a piece of cake.

[Splin]

Great stuff Floze. Very cool to see scene in different hours.

Synthesizer I think they are only half because of the way they were shot. In that movie they have, it says “Fisheye sequence” my guess is that these were shot with a 180 degree fisheye, which can only get half, or they used that half ball on the black table thing.

But still in pdf examples they show full 360 degree sphere shots where full sky is shown and on the edges you can see ground. Even in the home page header there are full 360 shots.
Oh well still usable in indoor lighting workflow what floze presented.

[floze]

Here’s a little flythrough movie of the 16-40 probe scene. Dont expect too much, I didnt tweak much since I’m not out to win a beauty contest with it (yet):

Divx5 compressed, 3 MB
http://www.floze.de/pics/cgtalk/skyprobes/tut_jap_ibl.avi

[Splin]

OMG Floze!
Thank you very much for rendering out ani. I love MR 3.4…or well, will love it if it arrives to xsi!
What were FG settings, same as mentioned above? Is FG map rebuilt in every frame or frozen(as in some corners flickering appears I guess it is rebuilt)?
thx!