Hi lexi,
The Fatten component is usually also my go-to for these kinds of things. The downside is that when you have lots of knots with various numbers of connections, things tend to get complex and messy. You could try to adjust the DirectionOverrides to get better results.
Maybe @DanielPiker could give you some more advice. If I remember right, he’s the developer.
Have you thought about simulating the inflation, instead of trying to get it statically?
This would be pretty straightforward with Kangaroo 2. You could start with a fattened mesh, where the staffs would have a uniform, minimum thickness (or diameter). Then you could remap your Perlin Noise values to define the inflation pressure at portions of the mesh, as well as other parameters used in the simulation. Here you have to plan ahead, how to identify the mesh portions to inflate!?
A lower noise value could for instance mean more dilation resistance or vis versa.
The simulation would also allow you to get these nice pillowy effects from Widrig’s example, which your fatten-only approach probably would not. Avoiding self-collisions are key here.
Furthermore, a by-product of simulation could be that some of the more problematic areas get relaxed or changed for the better.
Here’s a recent discussion about simulating inflation with an example: