3 ms·
Yeah, not just the real optimized motion but the simulated optimized motion looked significantly better than the original unsimulated animation. Makes me think
by ummonk 7y ago
Yeah, not just the real optimized motion but the simulated optimized motion looked significantly better than the original unsimulated animation.
Makes me think for actual animations they should simulate physical systems and then apply optimized control inputs to approximate the animation which the animator inputs. It would result in more grounded and realistic feeling animations.
- _Microft 7y agoThat might be because it's not longer too perfect. As far as I know little variations in timing are added to computer generated beats to make them sound better? That might be a similar thing?
- codeflo 7y agoTo make a different audio analogy: When they graph the input vs. the optimized control values, it looks a lot like the ringing artifacts of a low-pass filter. A low-pass filter removes high frequencies from an input signal. Since there are physical limits on how fast limbs can oscillate, maybe that’s part of what makes it look more natural.
- _Microft 7y agoI noticed that as well and it reminded me of the Gibb's phenomenon. https://en.wikipedia.org/wiki/Gibbs_phenomenon https://en.wikipedia.org/wiki/Gibbs_phenomenon
- ummonk 7y agoI suspect it's about reducing peak acceleration, jerk, and snap.