08-04-2024, 07:24 PM
(08-04-2024, 01:38 PM)josemendez Wrote: It is not possible to do this, since changing the softbody influence requires to recalculate and sort all per-vertex influences and then drop those above the desired amount of total influences per vertex. While I simplified this as lerp(skinning, simulation, influence) in another post, it doesn't actually work that way: full bone skinning and softbody skinning calculated, then their results interpolated per vertex. The reason for this is performance.
Say for instance there's a vertex that originally had 4 bone influences, with the maximum amount of influences being 4. Upon binding the softbody, there's 3 particles within the influence radius. So with a softbody influence of 1 (100%), the 4 bone influences would be dropped and only the 3 particles would have any effect, their weights calculated using the distance to the vertex and then normalized (so that all weights add up to 1). You end up with 3 influences instead of 7 (4 bones + 3 particles).
With a softbody influence of 0.5, all 4 bones and 3 particles influence the vertex, so they're sorted by weight and only the 4 having the most influence are kept, then normalized. You end up with only 4 influences, instead of 7.
With a softbody influence of 0 no particles have any influence and only bone influences are calculated, just like in a regular SkinnedMeshRenderer. Again you end up with 4 influences instead of 7.
This way the cost of the whole thing is not cost of bone skinning + cost of particle skinning: you only pay for the amount of influences actually used, for a typical speedup of x2. The downside of this is that you can't lerp between them in realtime, since the lerping parameter is "baked" into the weights.
Oh I see... that makes sense. Thank you for the answer.
I still think it might be worth it to provide this as an option for the user to use at their own risk though.