PDA

View Full Version : Characters - Which is better, simple or complex?



Teruchan
08-18-2009, 09:11 AM
I'm curious... I've been doing this for a very long time and speed has always been one of the most important things to me in creating 3D art and animation, regardless of whether it is cel shaded or not. Anyone who has seen Understanding Chaos or my other work knows I am mostly into anime style, but I have also worked on some films during my stint in Hollywood.

Lately, we know that sculpting and painting is all the rage, like that done in ZBrush, Mudbox or Silo. In my experience, that has led to some pretty heavy objects that can be a pain, if not impossible, for some systems to handle, especially if you don't have the latest greatest GeForce or 24X Core CPU. In my own work I have always gone for simplicity of the base cage, and allow for subdivision to do the rest, like this type of model here:

http://www.studioartfx.com/UltiBundle/files/blocks_image_21_1.jpg (http://www.studioartfx.com/UltiBundle/UltiBundle.html)

This allows the user to turn the sub-d level down to nothing if need be so animation can be quick, then shoot it up for rendering later. Of course, this is possible with the super sculpted meshes too, but it seems the base cage is always much bigger and heavier than I would like to deal with while actually working.

So which is better? Simple or Complex?

shrox
08-18-2009, 10:44 AM
...This allows the user to turn the sub-d level down to nothing if need be so animation can be quick, then shoot it up for rendering later. Of course, this is possible with the super sculpted meshes too, but it seems the base cage is always much bigger and heavier than I would like to deal with while actually working.

So which is better? Simple or Complex?

Would the sub-d method matter in this case? I haven't used the Catmull-Clark much yet.

jin choung
08-18-2009, 11:15 AM
the whole idea of zb and sculpt in production is to add high frequency detail at render time - so you have a relatively simple base cage to animate with and then a zb generated displacement/bump/normal map adds fine detail on your subd limit surface.

you do NOT use an ultra dense mesh for animation. that's suicide.

jin

AdamAvenali
08-18-2009, 02:52 PM
jin summed it up pretty well.

before the days of retopo tools, a lot of character artists would create a simple base mesh and take that into zbrush (or something comparable) and do their detailing and then export maps and apply them to their simple base mesh (exactly as jin has described).

with the recent advances in retopo-ing in programs like zbrush and 3d coat, a lot of artists (especially game artists) are skipping the simple base mesh and going right to the high poly versions and finally retopo-ing their high poly mesh into a simpler one.

so, to answer your question as to which is better, i find myself doing both versions eventually.

Teruchan
08-20-2009, 11:49 PM
What Jin said is precisely how I thought it was supposed to work, but when I was in the studio it didn't seem to be coming out that way in practice. It depended on the object, of course, but I remember lugging around some pretty heavy creatures in Layout and the speed of animation left a lot to be desired.

In one case, I think we actually used an entirely different proxy to animate with because the mesh we rendered was almost unusable.

jin choung
08-21-2009, 12:24 AM
In one case, I think we actually used an entirely different proxy to animate with because the mesh we rendered was almost unusable.

what you actually manipulate during production can vary. we're using the base cage for animation in my current project but in the previous one, we had to cut up the characters at the joints and then simply parent those parts onto the appropriate bones because the deformations were slowing things down too much (in maya, there's a free script called surgeon that takes your weighted character and does this automatically for you - so you can turn off the skinned mesh and just use the parented body parts as a visual).

especially if you have lots of characters in a scene, even the base cage can be too dense.

BUT... the point is - the base cage is the WORST CASE SCENARIO for animating (and perhaps more importantly) skinning, weighting, morphing and rigging.

you never EVER rig and weight the mesh's subdivided limit surface approximation.

jin

lino.grandi
08-21-2009, 12:43 AM
My workflow is LW/ZB based.

I always keep the cage object as simple as it can be (as long as I can animate every part of the model, that's fine).

Rigging and skinning are very easy to do on a simplified object.

Then ZB does the magic, adding any detail I need.

I'm currently working on some FX for an Horror movie (100% LW), with a lot of characters involved. Hope to post something soon.

pooby
08-21-2009, 02:50 AM
The thing about Zbrushed characters is that, if you just make a simple mesh and add all your muscles and detail with Zbrush. It will be the animated equivalent of a big rubber suit. All those muscles are baked into the UVs and
are just projected out from a simple cage, so will not look greatly convincing when moving.

This isn't what happens in good qualityFeature film effects. In a high-end character, Nearly all the definition of the character is on the actual mesh, and only skin detail (pores and wrinkles etc) is sculpted displacement. The mesh has to be of a high resolution to define the muscles even if the skin slides over them. (The definition provided by a separate underlying muscle rig)

Anyway all thats really a moot point as LW has no real way of making skin slide over a volume so you are stuck with the basic Rubber-suit technique.

jin choung
08-21-2009, 10:02 AM
Right, everything animatable has to have a representation on the base cage - but even still, that's a considerable diff compared to the rendered mesh.

Jin

Teruchan
08-21-2009, 10:59 AM
Boy, skin sliding over muscles is a whole different ball of wax. That sounds like adding calculation nightmares to an already time consuming process.

I remember some experiments a long time ago trying to get this to happen with the old Motion Designer plugin. Let's just say that didn't go well. ;)

pooby
08-22-2009, 02:55 AM
Its relatively simple to do in XSI, without any dynamic calculations.

but even if you dont use sliding skin, and just use morphs, you'd need enough muscle definition in your mesh and not rely on displacement for all that.

cresshead
08-22-2009, 04:05 AM
over in max you have cat muscle, in maya you have a muscle system also so lightwave it's going to be hard to achieve with no specific tools for mucles and sliding skin.

Teruchan
08-25-2009, 04:35 AM
over in max you have cat muscle, in maya you have a muscle system also so lightwave it's going to be hard to achieve with no specific tools for mucles and sliding skin.

I think that is a noteworthy point. It really is going to have to be specific tools for that purpose or it will not likely happen. Morphs are good and can be powerful, but they will have limitations.

http://www.studioartfx.com/UltiBundle/files/blocks_image_21_1.jpg (http://www.studioartfx.com/UltiBundle/UltiBundle.html)

I tried to put a decent amount of muscle definition in my model pack characters (http://www.studioartfx.com/UltiBundle/files/blocks_image_21_1.jpg), but I can see two entirely different results coming from using subdivision and some kind of sliding skin tool, versus morphs with what is available. On the other hand, I also heard that even some big movies have forgone sliding skin and muscle simulations in favor of morphs (or blend shapes) because they could get very similar results but at a much greater speed.