HDRI and GI setup in Brazil

[Aldaryn]

I think there are no" good and simple" GI setups adequate to all scenes in Brazil, like in every other renderer...
But I usually use regathered photon mapping to save time by QMC calculations. Some good (or good so far... :) rule of thumb might be to use 10-20 photons per every 50px of the rendered image. This amount should be eneugh for the raytracing to work without artifacts....
But if you simply use for e.g. an outdoor setup, it would be a waste of time to calculate regathering. Try to use 2 bounces with 15-20 view rays cast per smale.
Of course it is good to play with the sample rate too, -3,0 to start with is too high. In many scenes (if they lack hard edge transations) it is eneugh to use -4 samples per pixel...

About HDRI, you only have to drop it in the skyílight material slot, to illuminate the scene, no need for spheres. And use a slightly blurred HDRI image, if you do, you won't need too much view rays for the image to be "grainless". (Of course this will blurr shadows as well, but...)
Some good HDRI resources: http://www.debevec.org/Probes/
And you can also find a good HDRI setup tutor at www.splutterfish.com, rigth where you can download the I/O plugin for HDRI

Hope I've helped. :)

Man, this was my FIRST POST at CG-Talk. I finally got the guts to write in....
:cool:

[plaguelord]

Your first post and really helpful!

You can recomend some setting for indoor scenes?

[Aldaryn]

Thanks. :)
Regathering (setting bounces to 2, while using a global photon map) is the best for indoors, like in every GI renderer.
Also you will certainly need a high GI smple rate to render indoors without artifacts. Indoors usually deal with a lot o' corners and stuff... (modern insane architecture is deifferent.. yeah! :buttrock: )
I used to set it to 0,1 for production renders.
But if you don't have a lot of edges, and corners, use the default -3,0, and apply a bigger amount of view rays. Still some artifacts can appear, but saves a LOT of time...
Standard QMC raytracing is a big waste of time in indorrs, because you would have to use 8 or 10 bounces to get the same light distribution precision you would get from photons bouced 8 or 10 times around....
By the way, because Brazil don't supports photon attenuation (or light energy decay) you will certainly have to play with the exposure settings!

[ZeBoxx]

Originally posted by Aldaryn
By the way, because Brazil don't supports photon attenuation (or light energy decay) you will certainly have to play with the exposure settings!

I'm not sure what Brazil you're using, but SplutterFish's Brazil r/s supports all types of light decay and attenuation with photons as well as excellent per-bounce attenuation (which I'm thinking you're referring to) control.

Sections 5.4 and 11.3 of the manual should offer some insight :)

[plaguelord]

ok, let me see, sample rate 0-1, bounces 2, view rate about 20/30

can you explain what is Standard QMC raytracing ? :eek:

my scene have a lot of raytrace materials with fallof maps, and my idea is put my object in a table very close to a windows.

another question, in a 800x600 image, you have 480000 pixel, so I need 9600 x 10 photons?

thanks!

[ZeBoxx]

QMC Sampling basically refers to shooting out an N amount of rays (N = View/Secondary Rate) in a hemispherical semi-random pattern.
Apart from that, it is just like raytracing, in that when a ray hits something, it calculates the color with lighting etc., and if not, it'll calculate the background color (or optionally skylight color, in Brazil r/s' case).
Once you've got all the 'colors' for the rays, you more or less average them together, and that gives you the lighting color for the point you're rendering.

This is a fairly accurate method, but of course has various drawbacks.
high frequency / high contrasted results between QMC Samples and/or the results from the QMC Samples makes for noisy/smudgy renders.
Add to that the 'problem' that if you have more than 1 bounce, say, 4, it all works iteratively.
Assuming all rays will always hit and never die out below the threshold, and you've got your view rate at 32, then you're looking at 32*32*32*32 total rays - or 1,048,576 of them.
Not so speedy :)

-------------------------------------------------------------------------

Photons are a whole different topic. The number of photons you want to use is compeltely unrelated to image output size (unless you're rendering 1x1 or so). It is far more dependent on scene configuration, and I'd very much recommend going over our example scenes for this.

[Aldaryn]

ZeBoxx: I'm using the best around : Splutterfish's Brazil...:)
Maybe I've skipped the part in the doks, and judged by my experiences. With photon mapping my scenes seem sometimes a bit overlit, like at a longer exposure, even if I set up all the materials - I think - correctly, use utily mats. to refine GI distribution (e.g. a room, my latest project) and refer as close as I can to real world surface attributties... (e.g. again, the room, I've painted some planks with matte ang glossy paints to examine the spacular & diffuse light reflection, to have a good mat. setup)
After using photon maps the whole scene looks overexposed. Like the photons would maintain their intensity.
How can I solve that problem there?

[ZeBoxx]

So you have Brazil r/s... and you have the documentation... sounds almost like you would have access to the official support forums, too ;)

Anyway - the problem with photons in some scenes is that a fully white material will splat at 100%, and reflect the photon at 100%.
Obviously, no real world material is completely white.

You might want to use darker materials, either in cooperation with a utility material or not.

Like I said, it can definitely be done, but these things are best handled at the official support forums :)

[Aldaryn]

Thanks, ;)

By the way, like everithing, Brazil needs a lot o' experimenting, with the controls, and settings... and I think I'm getting along with my room very well lately... got some superb lightning done.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.