Wearables: conform to avatar

[avaer]

We can use the bone weights to vertices to estimate the size of limbs of an avatar, which would allow conformant wearables like Loot.

As part of editor features, we could add sliders on top of that for adjustments to a particular avatar.

Or perhaps the editor should come first?

[Vianvolaeus]

Sounds interesting. If the editor is for scaling / translating verts to make it conformal to the avatar's shape, then I would hazard a prediction that it should come first. Depends on how the weights are being calculated.

[plankatron]

i had a similar thought to this specifically for loot : here is the breakdown :

Loot Rigging

Problems

• vertex weight transfer is problematic - doesnt always transfer correctly small issues lead to glaring problems
• heat weight skinning / voxel skinning / voxel heat skinning all have problem cases and dont work 100% .. more like 60% , this is complicated with armatures that have extra bones not specifically related to the body like weapon mount locations, root bones, ik bones , leaf bones , other types of control bones
• shapekeys would prove to be too complex and tedious to implement, for every push out you will need a push in, bones are more explicit and each bone can be scaled/rotated/translated

Solution

• each loot item will have :
  • control bones for attaching to vrm's armature as well as control bones for controlling scale/rotate/translate
  • at minimum you would need a root ctrl and a child ctrl, root is parented to a bone on the vrm armature, child ctrl deals with rotate/location/scale
  • on top of this we also have a mesh of control bones for controlling the shape
  • from testing you need to usually have a minimum of 6 bones for all 6 directions, some may need more or less, this part needs testing per loot item category.

Conclusion

• this concept should in theory allow almost a complete arbitrary mapping of general loot items to general vrm armatures, even non humanoid avatars can be used.

[avaer]

@plankatron how is this going to work for the main use case (existing wearables where we do not control the data)?

Is the idea that we will have third party creators perform the above procedure on their assets? Or are we doing that ourselves on their assets?

In the former, what is the org structure that allows that to happen? In the latter, how long does that take for each third party wearable? Does it make sense?