Currently there are two scenarios:
A.Cloth for runtime gaming. Players could wear different cloths.
B.Cloth for game CG (rendered in offline renderer), but with a much higher efficiency than pure simulation.
The purpose A is solved simply by rig the cloth and skin it. Merge the cloth and character and then export. Game engines like UE handle this well and can add some proxy simulation to it, and players have a "acceptable" result.
However, for rendering a Promo-video, or major level cut scenes, we need a more visually realistic cloth. However we want efficiency as well because we don't want spend a lot of time tweaking the simulation. We know that vellum is fast, but we haven't test it with multiple meshes with more than 100,000 polys.
Some of the thoughts we have for possible problems:
P1.Character will have multiple cloths made from different physical materials, so we cannot merge independent cloths on the character.
P2.We may want to use external program like Marvelous Designer to make the mesh (because that's easier to use), and import it into Houdini. We don't know what if thickness matters.
P3.Complex interactions between character, cloth or even environments. Character might press and pull cloths of different physical materials, and cloth bend and distort differently. Wind force can blow up mage robes.
We are very sure Houdini can do all of this if done by expert users but as indie devs we need to find an efficient way to achieve (or fake) it.
My current solution for scenario B is like this:
S1.Character Rig&Skin; AND cloth creation in MD
S2.For hard cloth like body armor. Use character's skeleton to rig it and paint it with fixed 1.0 weight. For soft cloth like silk robe which is not containing the character (meaning it does not have closed topology), use vellum.
Now the problem here is in those "soft" cloth made of cotton, silk, and leather that contains the character, or cloths like a 18th centry royal dress. And let me call it "Jam Cloth" here in order to clarify. Jam Cloth is always containing the character, they always deform according the the deforming body, but their deformation has soft and floaty.
Therefore, rig it may not achieve our ideal result. If we don't rig it, we will expect vellum to solve cloth simulation AND multiple hi-res mesh collision (to make sure part of Jam Cloth follow body deformation )
Maybe we can use attribute paint to mask the simulated verts for a rigged cloth? So we can achieve effiency and quality at the same time? I don't know if it is an allowed feature, but by doing so I think we are "emulating Unreal style clothing sim"
What's your thought of the current clothing pipeline?
