>>2967588>It doesn't matter where the other vertices actually are.Hmm, I'm not so sure about that. You could go for "good enough", but you need to know the approximate vertex density for it to be exact. Here's a picture to illustrate, the same offset for the center vert, but one's vertices are closer to the center laterally. The smoothed normal is somewhat different.
But if you are going for semi-accuracy then yeah, I guess you're right, you can "sample" the approximation function at slightly offset points to get the slope. It might still be worth to use the ortho-camera-with-a-short-range-of-znear-zfar-distances-into-the-mattress so that you wouldn't have to use set bones to approximate the proximity, instead using rendered geometry to drive it.
