|
Ok, thanks for a lengthy overview.
One question remains though what has bothered me for a long time. What
kind of techniques are used to render out transparent objects when
rendering in passes.
Thanks again, both of you guys!
-T
peter boeykens wrote:
Ok, good to know. But the question still remains :/
Or I'll try to rephrase my question. What should I investigate to get
better understanding about the requirements of compositing right
layers in right order. Of course, before comping right stuff needs to
be output. Should I look to renderman world for it. Is there more
info available? Since everywhere in facilities guys are comping
together complex scenes. Where from do they gain that knowledge. Only
cause some guys in the " inner circle" know about it and are only
sharing it to other who also are in the "circle"?
there is no one right order or one right approach to this.
first off, CG is one big cheat of reality.
eg. ambient light, specular, shadow maps,... are simply tools that
vaguely resemble some reallife phenomena.
You'll have to have a good look at what you're trying to approach
(reference material!!!) , and then look to the tools you have at hand,
and find out how to get such results. Its common sense, but sometimes
people forget it also holds true for shading, lighting and compositing.
The look you are after will dictate to some degree how you should
light, render and composite it.
If you want an introduction into rendering in passes and compositing,
these are pretty standard:
http://www.amazon.com/Science-Compositing-Kaufmann-Computer-Graphics/dp/0121339602
http://www.amazon.com/Digital-Lighting-Rendering-Jeremy-Birn/dp/1562059548/ref=pd_bxgy_b_img_b/104-8706183-0187158?ie=UTF8&qid=1175162623&sr=1-1
there's various approaches to rendering 3D.
on one side of the spectrum you have trying to render everything at once.
Its how most people start out, and you quickly run into difficulties:
scenes get to complex to handle, certain rendering effects dont work
well together, rendertimes go through the roof, modifications require
full re-rendering, and you often dont get exactly what you're after.
Then again, I've seen people do very realistic stuff just in one go so
it can be done. There are still cases where this is the preferred
approach.
Next step would be to seperate between foreground and background, or
background and characters. It allows for several interventions in
compositing to make the characters stand out more from the background,
or inversely to make them more integrated.
A logical progression of this is splitting the scene up further:
backgrounds can consist of tons of elements, characters too. Crowds
can be split up. you can have far background, background, mid, and
foreground. You can end up with hundreds of elements. You can do this
for creative control or simply because the elements are too complex to
render at once.
And when there's modifications you often need to rerender only the
elements that are affected by the modifications.
This also raises the need of several utility passes, such as depth,
holdout or matte passes,... in order to layer the elements together
again.
A next step is to think about the "beauty pass" approach: creating a
good looking render in 3D, that you then further enhance or modify.
What exactly is present in the beauty pass and what needs to be
seperated is not written in stone. You might seperate shadows, in
order to optimise rendering, or for treating them seperately and for
finetuning them. You might seperate speculars, as they might require
optical effects or filters. And so on for reflections, refractions.
Soon you start thinking about splitting up everything. Every component
in your shader could perhaps be rendered seperately and composited
back together. And then you got the dilemma: how does the render
handle this internally and how to get the same result in compositing.
Truth is, there's no one-scenario-fits-all solution.
What happens internally is not just up to the renderer, but also the
shaders being used. If you would look at the code of different
illumination models you might see that some are doing quite different
operations. So as a result, splitting up into channels and compositing
them together again might look nearly identical as the 3D render for
certain objects, and be totally off for others.
So you might approach the issue differently: how would you like to
light in 3d and split up the render to get the most control and
flexibility in compositing?
This is a very big topic, and you could write whole books about it.
Some companies have established very precise scenarios and tools for
this, and sometimes such tools find their way to the public. (ambocc
comes to mind) But lots of artists have also developed their own
approach, often with great results.
Andy's breakdown somewhere else in this thread is very good.
But its by no means the one and only right one.
Just to give an example: some people use lights with their intended
color in 3D. Others render all lights as pure white, and define the
colors in the compositing. Some people like to render one light at a
time. Some like to split between indirect light and direct light.
Some like to render tinted shadows, some like to have black and white
shadows. Some like to use transfer modes (screen, add, multiply..)
when compositing, others use passes as masks for color correction
operations, or for mixing between different rendered versions.
This could be personal preference, this could be based on particular
lighting situations, or this could be based on the 3D and 2D software
used, or on the requirements of the production.
No need to go to renderman for this to answer your question.
XSI with mental ray is a very good platform for handling this.
Much as I'd like to go on and be more specific, it is such a broad
topic, you cant just cover it in a mail.
Even a tutorial wouldnt do.
It is something you pick up over the years, trying things for
yourself, running into situations where you need this or that, and
seeing how other people tackle certain problems. And right when you
think you know it, technology evolves and new approaches and tools
arrive. :-)
P.
Ok, good to know. But the question still remains :/
Or I'll try to rephrase my question. What should I investigate to get
better understanding about the requirements of compositing right
layers in right order. Of course, before comping right stuff needs to
be output. Should I look to renderman world for it. Is there more
info available? Since everywhere in facilities guys are comping
together complex scenes. Where from do they gain that knowledge. Only
cause some guys in the " inner circle" know about it and are only
sharing it to other who also are in the "circle"?
Andy Nicholas wrote:
Mental ray doesn't composite layers to produce the final image.
Mental ray
illumination shaders like Blinn only have one output (color), they don't
channel the components out to separate channels to be composited later.
Don't get confused by the user framebuffers. Shaders will write to
them only
when requested to. They don't form part of the rendering pipeline to
produce
the final image.
Andy
-----Original Message-----
From: owner-xsi(at)Softimage.COM [mailto:owner-xsi(at)Softimage.COM] On
Behalf Of
Tauno Ööbik
Sent: 28 March 2007 18:56
To: XSI(at)Softimage.COM
Subject: How MR cooks together final image
Where from can I learn more about the topic how MR internaly builds
together
final image. IE in what order does it calculate sub-passes, and in
the end
how does it put together all the layers together for final image. I am
looking that info to get better understanding about the internals of
MR and
in the end to use that info to put together more efficient passes for
compositing.
I already took a look to the well known book called "Rendering with
Mental
Ray" but unfortunately this topic is covered insufficiently there. Or I
missed the right spot :)
Unfortunately, right now I learned more about that topic from here
then from
that book:
http://www.jupiter-jazz.com/wordpress/wp-content/data/tr4kv2/html/chapter3-S
OFTHARD.html
I just wanna shorten my work days which are growing way tooo long as
the bar
needs to be risen again and again constantly.
Thanks a bunch,
-T
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in
body:
unsubscribe xsi
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in
body:
unsubscribe xsi
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in
body:
unsubscribe xsi
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in
body:
unsubscribe xsi
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in body:
unsubscribe xsi
|
