this is a old question but it was never really answered. and that’s annoying because it’s one of the first question that shows up on google. So I will give it a try.
Why a capsule and why not to change it.
Even if you have a quadruped character you don’t want to change the capsule to some other bounding volume. Even if it were possible. There is a good reason.
In a game that is mostly horizontal (as on the xy hyperplane) you always want a smooth concave shape that ideally has no “corners” so you dont get stuck in the piece-wise smooth and sometimes convex geometry with all kinds of edges and hooks that is the game world. If we cut the world in layers parallel to the xy hyperplane we observe that the capsule is always a circle and that’s ideal. It’s exactly what we want.
That doesnt mean that you will not collide with the colliders/ physics asset/ skeletal mesh of a clone of yourself and vice versa.
But I want to have a head or something to bump into walls and dont want it to clip
In that case you need to split your mesh into various actors and put the parts together with sockets and child actors. This could look something like this:
Moral of the story: use the capsule. Don’t try to rewrite the movement component just to realise it was a dumb idea.
edit (deprecated) : I only tried it so far with camera boom and in that case yes it works as expected but the collider of the attached mesh is ignored and only the collider of the camera boom is used. will find a better solution
edit2: Now we are getting somewhere. When you have a following set-up: capsule → spring-arm → childActor the spring arm will do the actual colliding and “move” the childActor towards the capsule until it collides. Yes the capsule will collide with the childActor.
Now, the question is how much sense does this make. The best move forward is to attach a bunch of springarms with decent sized probes and leave away the childActors altogether in order to “extend” the capsule.