tl;dr: I'm having some funky normal rotations on only a few line nodes (ones that are nearly aligned to the Z axis) after being wiggled around with a point vop.

I've been following along with various tutorials and combined a bunch of stuff together to make a wavy tentacle ball thing.

It's basically a subdivided sphere with lines copied to the center of the primitives. I'm using a point vop to give the lines some wavy motion. I am using a for-each to copy the primitive at the base of each line to the points down the line and skin it.

I wanted the sphere to be fused with the base of the tentacles, and tried polywire, extrude and sweep, but I couldn't get the control of the profile that I needed at the base. so I ended up with the copy/skin method. I have managed by some small miracle to get the whole thing working great except for the lines that are the closest to being aligned to the -Z axis. It looks like the polyframe is flipping the normals around funky. I've added the regular tangent/normal to normal/up vectors and turned off the up as it wasn't needed and didn't help the situation.

I've got the single pass turned on and isolated one of the lines that has the flipping. If the time slider is scrubbed a bit, it becomes more evident. All the other lines except the ones close to this one are fine.

Can anyone take a look and see if they can figure it out, it's the last issue and i've been banging my head on the wall for a long time with it.

Bonus question: I'm using a vopsop that I copied from an example to align the primitives to 0,0,1 and move them to the origin before my copy to points. This works great, but I tried to recreate it using a point vop and It doesn't seem to work correctly. The working and non working nodes are next to each other in my graph. Why is that?


Thanks



Edit: I was playing around with the point vop that I am using to give the lines some motion and if I change the noise to 1D then it goes away, but obviously the noise is not really that nice in only 1 dimension, so maybe it has something to do with how I'm applying the noise to the points?

Edit 2: I've attached a simplified scene that exhibits the same behavior.

So I used the Entagma parallel transport video to solve my flipping as the line approached the up vector, and it worked like a charm. Such a useful method!

Now, if I can just figure out how to get my first normal to match the primitive that's being copied. I tried to set it to the normal of the first point, but it doesn't match. If I turn off all the parallel transport stuff the copy to points matches perfectly, just with no rotations, so I know it can do it. One more puzzle piece…