Help Scaled SkinnedCloth not matching

[Jawsarn]

We have a setup of a animated character with a scarf. The scarf gameObject has a scale to it of 100.
If I leave the scale to default on the Skinned Cloth Blueprint the mesh gets sized down when running.
If I scale it up with same factor, it seems to be offsetted forward a slight bit and partly scaled up.

I can "solve" it by scaling up the model in the import setting by the same factor, but this solution will break once we want physics on another object to with a different scale.
Is this a known issue and is there any solution at hand?

[josemendez]

We have a setup of a animated character with a scarf. The scarf gameObject has a scale to it of 100.
If I leave the scale to default on the Skinned Cloth Blueprint the mesh gets sized down when running.
If I scale it up with same factor, it seems to be offsetted forward a slight bit and partly scaled up.

Blueprint scale affects cloth particles only, and scale values compound. If it's also offsetted, chances are the pivot of the object is not at the correct place - all object transformations take place using the pivot as origin, so if the pivot is to the right of the object, the object will move to the right side as you scale it down and to the left side when you scale it up.

I can "solve" it by scaling up the model in the import setting by the same factor, but this solution will break once we want physics on another object to with a different scale.
Is this a known issue and is there any solution at hand?

The usual approach to character cloth involves placing the solver at the character's root transform. This not only gives you fine control over inertial forces, it also automatically handles scaling (since simulation takes place in the solver's local space, scaling the character also scales the solver and the cloth simulation).

kind regards

[Jawsarn]

The usual approach to character cloth involves placing the solver at the character's root transform. This not only gives you fine control over inertial forces, it also automatically handles scaling (since simulation takes place in the solver's local space, scaling the character also scales the solver and the cloth simulation).

I'm not sure I follow here. We have the Solver on the root transform, and then one of it's children has a skinned mesh renderer which has a scale of 100, some rotation and offset.

The only way I've found to get correct default physics behavior right now is to scale up the mesh in the importer, and then inject a parent the hierarchy to the skinned mesh renderer with a scalar of 0.01. With this I did not scale the blueprint, and things move around with inertia.

A tangent question to what we want to achieve. Are skinned cloth more constrainted in movement compared to normal cloth? Our case of a scarf has danling parts, and I'm trying to figure out which solution (skinned, cloth, rope, rod, softbody) etc. Tangent to that tangent is that we want to have a character tail later as well, and similarly there are rope, rod, bone, softbody to choose from as well. Is there some general matrix of pro cons for each solution?

Best regards

[josemendez]

I'm not sure I follow here. We have the Solver on the root transform, and then one of it's children has a skinned mesh renderer which has a scale of 100, some rotation and offset.

Oh, sorry. So you want to scale a specific child transform of the character, not the entire character. What I suggested is to scale the solver, which will scale the character and all cloth actors inside of it.

and things move around with inertia.

You can control how much inertia the solver's movement injects into the actors it manages by setting the world linear/angular inertia scale parameters. This is useful for characters since it allows you to explicitly control how much the cloth moves when the character moves. For instance if cloth flails around too much when the character turns around 180º, reducing the angular inertia will result in less movement.

On the other hand, if you had a static solver sitting on the scene and the character moving within it, you'd have no control over inertial forces.

The only way I've found to get correct default physics behavior right now is to scale up the mesh in the importer, and then inject a parent the hierarchy to the skinned mesh renderer with a scalar of 0.01.

This sounds about right: if your cloth object has a transform scale of 100, it means its mesh is 100 times smaller than it should be so it needs to be scaled x100 to be rendered at a normal size. Creating the blueprint from it will result in a tiny cloth since blueprints are generated from a mesh alone, regardless of the transform hierarchy it will be used in.

By scaling the model up to a normal size at import time, you'll get normal-sized bones, a normal-sized mesh, hence a normal-sized cloth blueprint. But you'll still have an incorrect x100 scale at runtime due to its transform hierarchy, which you're "undoing" by adding a parent with a scale of 1/100: the children all have their scale multiplied by 0.01, so you end up with unit scale and proper mesh sizes.

You can't fix this by changing the scale value of the blueprint, since bones in the hierarchy will still have incorrect scales: scaling the physical representation of the cloth but not the bones used to drive it will lead to weird results.

My two cents: a scale of *exactly* 100 makes me suspect this mesh has been incorrectly exported, possibly from Blender: Blender's FBX exporter default settings will often lead to transforms with a scale of 100 and a rotational offset. Maybe it's not the case but if it is, the cleanest solution is probably to export it correctly. We also have a video on this subject. Having consistent scaling is quite important in general, as it keeps you out of headache-inducing situations.

but this solution will break once we want physics on another object to with a different scale.

You can have multiple blueprints with different scales, then skin meshes that also have different scales to it as long as the cloth and the renderer components aren't on the same object. You can even have multiple meshes skinned to the same cloth. See: https://obi.virtualmethodstudio.com/manu...modes.html

A tangent question to what we want to achieve. Are skinned cloth more constrainted in movement compared to normal cloth?

They can be, and they usually are. Skinned cloth makes use of skin constraints, which limit particle movement by keeping them in an area close to the mesh's skinned vertices. This allows you to blend simulation and animation, making sure the simulation doesn't stray too far from the reference animation. Most realtime character clothing engines work this way (Unity's built-in cloth also does), it's particularly useful for garments that are close to the body such as shirts/pants: using the skinned/animated vertices to constrain the simulation is much cheaper and way more robust than relying on cloth collision detection against the character's body.

Our case of a scarf has danling parts, and I'm trying to figure out which solution (skinned, cloth, rope, rod, softbody) etc. Tangent to that tangent is that we want to have a character tail later as well, and similarly there are rope, rod, bone, softbody to choose from as well. Is there some general matrix of pro cons for each solution?

Rope is a 1D chain of particles: best for things that are long and thin. Particles have no rotation (they're just points), so the rope can twist around its length freely.
Rods are also a 1D chain of particles, but they do consider orientation: You can control how they twist, and give them a specific rest orientation.
ObiBone is a particle tree , created from a bone hierarchy and driven by its animation: best for pre-rigged stuff like tails, braids, limbs, etc.
Cloth is a 2D net of particles: best for things that are mostly flat and have little or no volume.
Softbodies are a 3D network of particles: best for things that have some volume.

Which one to choose depends on your use case's specifics. So in case of say, a scarf, which is a long and relatively thin piece of cloth: you could be tempted to use a rod, rope or bone (softbody wouldn't make much sense since a scarf has basically zero volume), but the problem with these is that a 1D chain of particles won't orient itself properly when colliding with a surface, it may roll and the scarf may end up sideways on a surface.

You need at least 2 rows of particles for it to sit correctly on a surface, so the obvious choice is cloth. Poor diagram I drew, top half illustrates a rod/rope scarf colliding against a plane (may rotate and sit incorrectly on a surface), bottom one a cloth scarf:



kind regards,