Lighting, Compositing, Setup Guide
July 6th, 2008
For the junction environment I’m just going to share a bit the lighting set up, the passes I’ve used and the way they composite together.
If you want some more information about the passes I use, why and how they are composited togther, check out my compositing guides on my site here: Part 01, Part 02.
As a start, here is the compositing tree in XSI’s FxTree.
In the top left you can see the beauty pass, then following down from it the different ID passes and the colour corrections using those mattes. Then on the right there is mostly the depth, the background and the ambient occlusion with varies grading.
In the (bottom) left area there is some extras like glows from the street lights and the sky.
On the next series of images the node which is shown in the image is highlighted on the tree. Click on any of them to view a larger version.
The first image is the beauty pass straight from the render. The dodgy pixels around the edge of the sky/buildings, is the alpha channel that has been un-premultiplied and gets that sorted back again in the ‘over’ node when the background is added.
There is no GI or final gather etc. in the renders. The beauty pass is made up of infinite fill lights, and additional rim or specular-only lights which only affect specific objects too add some highlight and shape to some of the objects.
For instance the barriers on the islands in the middle, the church at the back, the traffic lights, and the bits on the near right side - they all have some additional lights, some just spec, some diffuse as well.
The following are the ID passes (at the moment) that are used to grade certain areas of the image, mostly just on the beauty render, but they can potentially be used at any point in the tree.
And the depth pass, this one is aliased as it is used in the comp, but also I render a version with no AA, which may also be a different range of distance from black to white, and at a very high resolution for the depth of field if there is some.
Ambient Occlusion which is then tinted and graded slightly (changing of the levels)
Using a multiply node, the AO and the beauty pass come together.
This is the same depth path, with a change of levels, and graded/tinted to be used as haze/fog type stuff.
The haze added!
The background, which is from a photo I have taken, but also painted up a bit, the building on the left is only still there as you cannot see it 
The passes is merged over the background, notice that the dodgy alpha edges have gone now and we don’t get any artefacts, dark alpha edges around the buildings because of this process.
Grading, grain, and some nice glowy bits brought in from other passes are added and we end up with the final image.
Not too in-depth but I hope it is interesting to see some of the passes and a few techniques.
Cheers!
Nice post! I’m putting the finishing touches on the next podcast and one of the things it talks about are those RGB matte passes. I’ll definitely link to this when its up. The one burning question I have is the depth pass. It sounds like you’re saying that you get better results when doing DOF with a higher res image? How does that work exactly? Does it just handle the edges better?
Comment by Lucas Martell — July 6, 2008 @ 4:47 pm | Quote
Great compositing job, I really like how the final image turned out with final color grading. It is the first time I have seen any serious compositing in XSI and I’m glad to see how powerful it is. Although we use XSI for our projects I never really got the hang of using it’s compositor and we mostly use Fusion for our projects.
I was wondering whether your AO passes take a long time to render and if so, whether you have ever considered baking the AO into the texture or creating one huge AO texture for the AO pass. I tried to find such a solution but I always bumped into UV issues. Any thoughts on the matter?
Good luck with the rest of the project which is turning out to be a feast for the eyes.
Reza
Comment by Reza Ghobady — July 7, 2008 @ 6:44 am | Quote
It depends a lot on the shot, I think with the one with Edgar in the street it has been quite a big transition between focus and out of focus so its been tricky.
I basically did what was is explained in the Lenscare Manual, which is rendering the depth 2x (or more infact) then resizing the comp to match that, then DOF, then resize back down and the edges should match a bit better.
If you use dof with a anti-aliased depth pass all the edges that have been smoothed grey between white and black, and so for those wee pixels it will be set at the wrong distance.
Comment by Simon — July 7, 2008 @ 1:45 pm | Quote
Thanks for your kind words Reza, the actual AO doesn’t take too long really, but I have found the pre-calc part does take a while (I can’t say at this point if rendering more than 1 frame in one go, it will do that slow pre-comp every frame). About the baking though, have you thought about camera mapping? I have been thinking about that myself for the street scene.
In the end it might not really matter if I just let shots render away while I’m at work during the week :). But perhaps camera mapping the AO only might be a good idea. Certainly better than baking in my experience.
Lucas talked a bit about it in one of his podcasts, that might be handy.
Comment by Simon — July 7, 2008 @ 2:09 pm | Quote
[...] has a great two-part compositing lesson (part 1, part 2) and also a work in progress showcase for Bank Affairs showing some great compositing [...]
Pingback by cg*Pipeline » Compositing - Prologue — July 19, 2008 @ 10:25 pm | Quote