> I'm curious if the algo is fast enough for real time or if it's only for renders.
Yeah, likewise. It definitely looks like it's formulated for optimizing paths rather than, say, simulating real-time physics, so I'm guessing not yet. From a quick scan of the paper, I didn't see anything on the complexity (O(N logN) or O(N^2) or whatever), but there were some timings:
> Figures 11 [clasped hands] and 15 [underwear model] each took first 10 minutes to find a coarse trajectory, then 5 and 2-3 minutes, respectively, to refine [to 60 fps].
Those numbers don't mean much beyond "minutes are better than days", because this contribution is a quality improvement via a new algorithm. Optimizing performance and maximizing choice of hardware power is another effort.
Is there a standard underwear man model like the teapot? I find it's use hilarious considering there's no reason for model to be in underwear. Same with the butt/hip model at the end. Chefs kiss.
The main example is a dude moving his arms around, and the repulsing shell/surface kind of interpolates through the motion semi smoothly.
I could totally see something like this being useful or interesting for destructive physics in video games. Like, oh, this surface/shell is about to repulse a rocket slamming into it, deform it for this hit. It's be a very different application but there's a kernel of morphology here feels similar ish, of this kind of softbodied-at-a-distance simulation.
Based on other comments though it sounds way too slow to consider for realtime games though.
Does this have implications for gaming? I know character models with lots of "accessories" clip into walls and the character body very often, and from what I gathered, the belief was that this was a hard and manual problem to solve.
Might speed up the animation pipeline so artists can get to something physically plausible quicker but it doesn't seem like it's feasible to run in real time.