View Full Version : Aneks multipass tut

11 November 2004, 05:58 PM

Really incredible tut on multipass. Thanks so much. I'm just getting started in Shake so have some questions hope somebody might be able to answer

1) If the shadow pass for the man had been done separately would it just have been added in with a gamma like the other shadows after the keymix and switchmatte are used to add ambient and diffuse colors to the man or would it be brought in before the diffuse_man is brought into the switchmatte (or does it make a difference)?

2) Why is the shadow for the ball in the g channel to begin with. Is that how it was brought out of the 3d package and if so why?

3)Why is the screen layer of the ball then sent to a switchmatte node with the ball_diff as a matte. Don't they have the same alpha? Is that to carry out premultiplication? The same thing is done with the floor so the same question there but I also notice that the floor_diff being brought into the switchmatte after the shadows are brought in actually shrinks the alpha slightly Whis is the floor-diff alpha slightly smaller than the alpha coming out of the gamma and into the switchmatte?

4) Related to #2 above the red and green shadows on the floor. When he says it was rendered this way--is that something done in the 3d programme or prepreparation done in shake?

5) Why is the ambient occlusion reordered and sent to the alpha channel also?

Thanks for any comments. I found this an incredibly useful tutorial. Sorry for picky questions but I really am trying to understand every detail.


11 November 2004, 07:42 PM

Couldn't figure out how to delete this second post so just edited it. Thought I''d figured out the answer to one of my questions but in thinking about it later realized the answer made no sense.

11 November 2004, 11:47 PM
oooh someone read my tut ! I feel so special !

Thanks for the kind comments !

I wrote this a tut while back so I am working from memory here !

some answers

1) does make a dfference but not a huge one. In this case I would have used a gamma after the switch matte, because this is an 8bit example and once I had determined the ambient/diffuse shading relationship I would then go on and apply shadow pass to the result.

2)These shadow passes were generated out of Maya using Mental Ray and a trick I posted a while back on the maya forums which basically uses shadow colour and some very simple settings to get lights to cast pure saturated shadows in one colour. This means that one can get a shadow for multiple objects, seperately in a single rendering pass saving lots of cpu tme ! why green just a nominal channel could have been any one.

3) switch matte was for premultiplication as I find it confusing sometimes to use the pre_mult on layer nodes etc .... if it goes wrong you have to go looking for where the premultiplication is occouring and it can be a pain in the butt

for the floor I can't quite remeber this bit ..... maybe beause the shadow pass had a slightly dodgy alpha at the edge of the floor. ie it was shadowing itself !!?!? not too sure will have to check !

4) as in answer 2 yep ! A bit of overkill in this paricluar exaple but it is a general practise of mine to try and seperate out shadows as much as possible for more control ! As a side note I got this technique from an oldschool compositor who worked on final fantasy. He explained how the "ghosts" were created using a technique based on isolating certain 3d elements to channels and then colouring and filtering them in 2d. If you look at the special features on the dvd you can see the extensive use of rgb channel passes on both the ghosts and the explosions for the show !!

5) because I used an iMult and this node uses the alpha value of the background image to multiply with the fg. The dirtmap shader produces identical occlusionresults in the rgb channels and a traditional alpha. in order to get thr grayscale image to multiply the fg input on this node, i need to move the relavent data from one of the rgb channels into the alpha and then plug this into the bakground input

hope my answers make sense.

11 November 2004, 10:01 PM
Hi Aneks,

Thanks for the detailed response. Answers make total sense to me. You mention other tutorials you'll be working on at some point. I'll keep an eye out. My goal is to learn Shake (working on combustion too but Shake is really helping to open my eyes to the concepts behind what's being done) and try to get work in a compositing department so I'm chewing through everything on Shake I can get my hands on. Thanks again.


11 November 2004, 10:15 PM
Yes more tutes... Today is the last day of my contract in my current job so hopefully I will have some time on my hands very soon. I am hoping to get the motivation to put all these bits anpeices of informatio I have lying around into proper tutorials then !!!

11 November 2004, 09:45 PM
Sorry if this might sound stupid, but can anyone tell me where to find Aneks's tutorials?

Thank you in advance,
Tassos Ringas

11 November 2004, 11:06 PM

Here are the two tuts kindly being hosted by vfxtalk, at the moment. post me and hassle if you want more !! (

11 November 2004, 07:23 PM
Thanks Aneks! Beautiful tutorials, and although I am using DF there are very useful.

I have a question: what does the Re-order and Expand nodes do in Shake?

Tassos Ringas

11 November 2004, 05:52 AM
thanks for the kind words.

reorder is almost the same thing as digital fusion's channel boolean operation. It allows you to switch and substitute any incoming data from one channel to another for output.

When I refer to "reorder it gggn what I am saying refers to the normal order of the "reorder" node in shake.... which is r b g a z (people don't often use 'z'). in this case the data from the
green channel of the input image is replacing the data in red, green and blue channels of the output/result. Setting a channel to 'n' means the value of this chnnel will be "nothing" or 0. This is the not the same as the 'do nothing' setting in a digital fusions channel boolean.
Channel boolean is probably a bit more versatile.

expand / compress in shake do exactly the same thing with an inverted math. Basically they allow you set the range of values of the data in any channel thereby lightening, darkening or making an image more red for example .....while still maintaing the distribution of the values in that channel. This operatiopn is often confused with "clamp", but clamp merely sets a min and max value and disposes of the data outside the range.

Expand is my favourite colour correction node in shake as you can pretty much do anything with it. As I dont have DF on this machine (i am interstate at the moment) I am at a loss to think of an equivalent right now.... just use the awesome colour corrector tool in fusion and I will post an update when I fly back into sydney

hope this helps !

11 November 2004, 12:00 PM
If you're interested, I'm writing a plugin for DF at the moment that includes a lot of 'basic' colour correction nodes, instead of making you use the 'all in one' CC node...

Things like Add, Mult, Expand and Compress....

CGTalk Moderation
01 January 2006, 09:00 PM
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.