Hi!
Im looking for some insight into how character creator´s like the one in Skyrim and similar are setup and works together with animations.
Mainly Im looking for a good solution for faces, how the bone-setup would be and how the customize bone-setup will work with the animations for the face.
Anyone know of any good tutorial, GDC talk or such that touches this subject?
Feels like alot of games have done this and I can see how it can be done but it would help alot to suck up some experience from someone who have already done it.
Regards
Replies
Yeah they sure have alot of rigging tutorials but I cant seem to find anything about custom character builders / character creator, like the one in Skyrim or alot of mmos/rpgs.
Im looking for something like the stuff Naughty dog releases, example
https://www.youtube.com/watch?v=myZcUvU8YWc
GDC 2011
Ultimate Customization in Character Creation - Challenges and Innovations
by Brad Stokan (Cryptic Studios)
www.makehuman.org
helps?
In game you will just have a nose slider which blends between the two morph targets. Once you are satisfied with the appearance the game will bake the final mesh from the blended morph targets.
Blending is just a linear composition and it affects vertex posion, color, bone weights:
final_position = alpha * tiny_nose_mesh_position + (1.0-alpha) * big_nose_mesh_position
Alpha represents the slider for the nose. This could be done with mulitple morph targets.
If only I had the possibility to use morph targets
Cant say that I read all of www.makehuman.org but I scanned trough it but could find any, thanks anyways!
What Im lookin for is something that can give me 100% skinbind to two bones. One bone for customizing and one for animating. How were the character creators done before blend shapes?
Anyone know that sollution TES:Oblivion uses? They prolly didnt have blendshapes.
Right now Im using 50% bind to each bone but I think the animation is going to suffer in some areas.
http://www.facegen.com/
noseadjust.scale = scale * slidervalue
You could do the same with position to move it around the face and push it up and down too. Blendshapes/morphs are really going to give you a better result though. Most games will do a combination of the two I believe.
Once the final mesh is designed in the editor, it is better to bake the result in a single mesh and render the baked version instead of using some morph-target solution in-game. This way you could code the morph-blending by hand (CPU), bake the mesh (CPU), upload the mesh(GPU), which should work more or less with every engine.
Blendshapes are much older than rigged animation in videogame history, in early 3d games vertex blend was the preferred way to animate models. Rigged animation was a lot more expensive until vertex shader took the work of skinning away from the CPU.