What I thought would be a fairly simple setup for the new hair collision has instead proved very aggravating. The task is simple: have a character’s shoulder collide with her ponytail as she turns her head to one side. Dozens of attempts with various hair & collision settings, different key counts in the hair, and different hair weighting, yielded absolutely zero collision. So I simplified the task to test it and finally achieved some collision but it then fails within 20 or so frames:
Frame zero shows the basic ponytail hair system (see the weights image below for more info). At frame thirty the collision finally works and the strands at least sit on the back of the shoulders as expected. But then at around frame 50 the strands start to fall inside the collision mesh until at frame 110, all collision has failed completely. Any idea what gives?
This is not the original weights configuration, I lowered many weights so that the hair would fall away from the ponytail configuration and settle on the shoulders, the least difficult collision problem I could imagine. But notice above that some of the strands do not react to the Hair Dynamics AT ALL, they simply stay in the original configuration. I checked the weighting multiple times, repainted it half a dozen times, and made sure there are no hidden strands, and you can see that the keys are all colored as they should be, but both the Hair Dynamics and the collision seem to fail rather miserably.
I’ve seen a number of videos where the hair collides well with a moving object in a simple test setup, but what good is it if it cannot operate in a normal, typical animation situation?
Other factoids: My model scale is 1BU = 1m, the same used for physics calcs as I understand things, so a collision distance of 0.20 or more that I have seen recommended is impractical, as that equals 20 centimeters, and in a normal human, most hair falls well within that limit when at rest, making the collision calculations worthless (the hair must start out outside the specified Outer collision distance). So I tried much smaller values and at about .003 (3mm) I started to get collision. I had to max the sampling rate to 80 and put my subsurf mod above collision to get it to work fully, vastly increasing bake times, but then it goes and fails after a handful of frames. Hair collision was never this faulty when it used SoftBody physics (e.g., Kata, in my sig), why is the version using Cloth physics so cranky & unreliable?