They're all SUB-D modeled with proper support loops, a few smooths in and some nice detail maps and bam, you got your model last I recall, unless something is very organic and can have displacement maps on it like a dinosaur of the squigies.
Whoa that's an old one We've been just testing Zbrush in production back then, and besides we only had 6 weeks to deliver that 30 seconds, so there was no room for any second thoughts. That Tzeentch champion or whatever relied on displacements mostly, with very little facial animation.
Nowadays we go much higher, the average character (the control cage, before the render time subdivisions) is anywhere between 50 and 500 thousand polygons. It's usually around 100k, with 10k for the heads only to better accomodate blendshapes for facial animation. Of course it also depends on the character design, chain mail takes a lot more to model than a naked torso.
We also try to maintain even poly sizes so that when the renderer subdivides and displaces, the details are also evend out. We use normal maps too, raytracers can't handle subpixel sized polygons and displacement at reasonable speeds (Blur uses Vray, we use Arnold; Blizzard's on PRMan though but I dunno how much they raytrace).
There are also UV smoothing issues across the different apps - Maya, Zbrush, Bodypaint - which are best solved with nice even poly distribution, too.
For hard surface stuff I usually browse the related threads on this forum and the mantra is 'more geo' Had to do some armored character stuff recently and it helped a lot. Can't talk about the character yet but it was about 160-170k in the end.
One day I'll try to get the permission stuff done for some recent characters... Or you can look up blur's stuff, it's quite the same really.
In the article I posted, Blur resorted to using Mental Ray after a small struggle internally, for Hellgate too last I recall, is that not correct information Varg?
thanks, its good to know,
its been a while I haven't learn some cinematic pipeline. gotta need to update my knowledge a bit !
some last year and this year cinematic just blew my mind. the character expression became so alive, i guess the facial can be very high poly due to complex movement from the wrinkles
Replies
Here is the full thread for Warhammer Online from Blur artists: http://www.zbrushcentral.com/showthread.php?t=62638&page=3&pp=15
And yes, that's how it's mostly done, I think even Blizzard follows this pattern last article I saw.
Nowadays we go much higher, the average character (the control cage, before the render time subdivisions) is anywhere between 50 and 500 thousand polygons. It's usually around 100k, with 10k for the heads only to better accomodate blendshapes for facial animation. Of course it also depends on the character design, chain mail takes a lot more to model than a naked torso.
We also try to maintain even poly sizes so that when the renderer subdivides and displaces, the details are also evend out. We use normal maps too, raytracers can't handle subpixel sized polygons and displacement at reasonable speeds (Blur uses Vray, we use Arnold; Blizzard's on PRMan though but I dunno how much they raytrace).
There are also UV smoothing issues across the different apps - Maya, Zbrush, Bodypaint - which are best solved with nice even poly distribution, too.
For hard surface stuff I usually browse the related threads on this forum and the mantra is 'more geo'
One day I'll try to get the permission stuff done for some recent characters... Or you can look up blur's stuff, it's quite the same really.
its been a while I haven't learn some cinematic pipeline. gotta need to update my knowledge a bit
some last year and this year cinematic just blew my mind. the character expression became so alive, i guess the facial can be very high poly due to complex movement from the wrinkles