View Full Version : Animation / compositing question

Alex Rooth
12-11-2003, 02:55 AM
I'm not sure if this is the place to ask this as it is fairly general, but anyway:

Last night on the BBC there was a program tracing the development of animation from the early days to the present, concentrating on the development from 2D to 3D. What interested me was the shots of animators working on Woodie from Toy Story and a fish from Finding Nemo. They were tweeking facial expressions. What I found interesting was that they were basically in front of blue screens. Does this mean that when animating films like this, the characters are animated separately and then composited together with the rest of the scene? i.e. not rendered together with the background?

To make the question more specific to these forums, is this that people would work with DFX+ and LW?


Alex Rooth
12-11-2003, 02:58 AM
I meant is this the way people would work with DFX+ and LW?

12-11-2003, 05:45 AM
Dunno there - i guess that the characters would have to be part of the scene to cast/receive shadows and to be computed into the lighting/radiosity/ray tracing calculations.

They wouldn't need to be in front of a blue background either, because the 3d software would generate it's own alpha channel to composite with (rather than blue screening the character into the background).

Maybe it was just a draft stage they were showing, where the background hadn't yet been built, or it was taken out to speed up the responsiveness/cut out distractions to the animator - then the blue background was there as a neutral background?

I did tape it last night, but i haven't watched it yet...it sounded good. Yentob's series isn't it?

12-11-2003, 05:49 AM
of course, sometimes scenes are rendered as multi-pass...i.e if processor speed is at an optimum and there are lots of detailed elements to it, sometimes it get's broken up into chunks and sent to seperate rendering clusters. Then it gets composited together again, to cut down on render times and the intensity of the scene that has to be rendered.

For example, a big space battle, where you may have lots of spaceships to render, or a better example, a forest of trees, each one having it's own leaves and branches in motion. Then it would make sense to split it into foreground/character, midground, and background.

12-11-2003, 07:15 AM
Hello Alex.

It's a great way to use LW and DFX+, the amount of control you have in post production far exceeds what can be done in LW alone.

eg, You render a scene.
You have a city backdrop
Lights within the backdrop
Cars in the streets
Smoke smog etc
People walking on the pavements
Neon signs glowing above shops/walkways
Two main characters talking in the foreground

If you were to render all of this in LW (or any other 3D package) not only would it take forever, but once it's done, it's done. Other than changing basic contrast, colour effects etc there's not a lot more that's realistically possible.

Rendering everything in seperate passes - light layers, shadow layers, character layers etc, everything can be seemlessly blended together afterwards, and you still have control over the individual layers.
Blur the foreground or progressively blur the background, increase the intensity of the neon glows, render more people on the pavements and add them to the mix, Change your depth of field between your two main characters as they talk, take out the rendered city backdrop and add the real thing...

It's up to you.

For a good insight into how a large 3D production is composited have a look at
'Final Fantasy - the spirit's within' (4hrs extra boxset)

It gives a pretty good idea of how they built up their seperate layers, if memmory serves, they were using about fifteen seperate layers to construct the scenes.

Have fun :)

12-11-2003, 07:35 AM
Hello again,

There is another reason why large production houses build up their scenes in layers.

If one layer doesn't satisfy the producers or clients wishes, it can be rebuilt and re-rendered far quicker than re-rendering the entire scene - a massive cut in production costs.

Money, Money, Money. ;)

Alex Rooth
12-11-2003, 01:37 PM
matd: yes, it was Yentob's. It was okay, though quite general. Some of the Nick Park stuff was interesting, I thought, in the contrast it gave to CG - the way they deliberately leave thumb prints in the clay, that sort of thing; they are just down the road from me as it happens. Like you, I assumed characters would need to be rendered within a scene to get proper lighting, etc., but I'm just guessing really.

TLight: thanks for the FF tip - I will have a look at the DVD. Compositing seems to make a lot of sense. I really need to look into it.


12-12-2003, 02:37 PM
Pixar only just recently started compositing with any regularity.
So the shots you saw were probably not ment for compositing.
Mosy likely everything was just turned off in the scene. It saves time in the feedback. Also the animators have nothing to do with the creation of the sets or the placement of the cameras. The blocking has been layed out and approved by the directors by that point as well. Unless there is going to be interaction between the characters and an object there is no need for any of that to be there. It speeds everything up.
Oh yeah if it was facial expressions you were seeing. It was a face cam set up.
you can do this with lightwave as well.
just parent a light(that is turned off) to the head or head bone. point it at the face and there you go. in one of the windows just have it show the veiw from the selected face cam.

As for the compositing thing. You would be suprised how little of it is actually in those movies from Pixar and PDI.

12-12-2003, 04:54 PM
Well I'm working on my first animation with still real plates (back ground photos) at the momment and it's characters so basically this is how I'm working, NOTE I'm working, may not be the best way of doing things, just works for me. After using this I'm sold on it and use it for total 3D scenes as well. Not just merging live and 3D.

Basically load DFX with the back ground photo set to the size have a good look for shadows lighting set ups etc..(my pics are all indoors)
Load up lightwave setup the character animtion then camera lens size to match the scene and then set your lighting as nears as.

For instace I have a character walking down a corridoor (not well lit) in photo to into an open area which is well lit so as my character goes forward he get lighter.
Set this up in the scene render it out in 32 bit (no need for the blues screen)

In DFX create a new loader to merge with the background plate play it back once then in RAM it's playing in real time have look see what you get, typical additions are birgthness/ contrast altering the colours (all extra nodes to the flow) etc etc..
Some time once I've seen it I'll go back maybe alter a walk cycle or maybe the character gets too light too soon so I'll just adjust the light rerender back into DFX unusally just delete the second loader then recreate it (the loader and the rendered images are pullling or sending to the same directory) then rerender in DFX in RAM and see the results.
The main thing which is great is DFX all the colour adjustments etc stay the same so all I'm doing is using differnet loaders on each scene for the back grounds and the animations.

like I say works for me. Each character I'm rendering out indevidually as well so if I need to alter something with one I'll only go back and rerender that one and so cutting down on render times. as others have already stated. :)