Facial animation: Blendshapes or bones
polycounter lvl 18
Question being at your studio/with your engine which is used for facial animation, bones or blendshapes/morphtargets?
I'm assuming bones, as thats what the impersonator plugin/module for UT2k4 used, and thats what was mentioned as being used for Ratchet and Clank in an old gamasutra article, but most or rather all of the online resources and rigs for facial animation I've found rely on blendshapes. So I was wondering which is the real deal for game art.
I'm assuming bones, as thats what the impersonator plugin/module for UT2k4 used, and thats what was mentioned as being used for Ratchet and Clank in an old gamasutra article, but most or rather all of the online resources and rigs for facial animation I've found rely on blendshapes. So I was wondering which is the real deal for game art.
Replies
At the end of the day, I'd say go for morph targets if you want to do full-on lip-syncing. If you just want to give a character's face some different emotional expressions, go for bones. (they are easier and faster to rig up than morph targets)
Pro's:
Bone's can be blended just like morph targets.
Bone's are faster to setup and pose.
Mesh deformation can be handled by the GPU.
The CPU only needs to calculate a few bone transforms as opposed to hundreds of vertex transforms.
Programmers don't need to design/work/debug a morph target pipeline and animation system.
Con's:
If you don't have enough bones on a dense mesh you can't get the right details.
Some of the finer details you can't do unless you dedicate a bone specifically for that area.
Bone's are faster to setup and pose.
[/ QUOTE ]
In my pipline this isnt the case, considering that I am using the same vertex count and mesh topology for head for every head. Because of this I only had to set the blendshapes up one time and they are translated to every head in the game in one shot. If I were using bones I would have to rig and weight every head. As far as the posing. I have a interface/workspace (as its called in FaceFX) that I use to set up facial expressions very quickly.
Bones can be automated to an extent, so if you are using lots of heads, you can use the same animations and rigs and modify them, whereas blendshapes will have to be sculpted for every mesh.
So like everything in games, it is a tradeoff.
...If its 50/50 I suppose it would behoove me to have samples of each in a portfolio.
[/ QUOTE ]
Nah, just pick one and do it good. If someone asks about it in an interview setting, say something clever like, "I used bones in my example, but I also have in-depth knowledge of morph targets."
My own experience has shown that morphing works out considerably faster to create, considerably easier to manage in a pipeline and usually considerably better looking.
of course
yes..