Why aren't people looking at LA Noire as the future of facial animation?
Quote: "And animations are probably atom assignments so that's fine."
The problem is that the more vertices a high poly object has, the more processor time it takes to calculate the frame. If you have a main character with 10,000 vertices then that's fine. If you have a character with 1,000,000 points of information, well that takes up processor time. So much processor time, that I have no idea if it's even possible on current architecture, without dragging performance down to insufferable levels. Now graphics hardware can calculate vertex positions quite efficiently, so most of that drudge work is done on GPU's these days. UD does not have that luxury.
They say that you can't measure the properties of things like atoms effectively, because in doing so you affect it too much. This has similar problems. Displaying these models is one thing, because it can be indexed and then searched really quickly, making it possible to render in realtime. But - this all depends on the index remaining constant, and with animation the whole index would have to be dynamic.
The animation system would have to compensate for the amount of calculations, maybe for each point, add in a flexibility value, and then points allocated to a single joint could be calculated faster, no blending the calculations between joints. That's one thing they could do to maybe make animation conceivable, but then I guarantee that the quality on polygon model animations would far exceed it.
Now that is my opinion, I don't believe in magic, the tooth fairy, santa, elves... or the pixel fairies who come along and promise the earth but fail to provide suffucient evidence.
Health, Ammo, and bacon and eggs!
