Looks pretty damn nice, would have preferred to see more exaggerated examples (as well) for the sub surface scattering. I kept looking at his ears for the most obvious sign, but couldn't spot it there.
Funny that nVidia is using a technique that AFAIK ATI first used, that is, using RTT and then blurring them to simulate SSS, there is an ATI paper on it/skin rendering.
As for translucency, such as behind the ears, what they are doing doesn't cover that.
This is just a really high-resolution texture (4096 I would guess) and a really high-poly model. I normally love these sorts of demos, but there is nothing new here that a semi-competent shader programmer can't do already, using techniques and samples from shaders that have been out for a while now.
Alright, I just read through the 120-page nVidia paper on this, it is basically the same thing but improved (they mention this in the video). There are some cool things here but nothing new, or groundbreaking, or outside of the box... they're just doing what has already been done but was too expensive to do (6 specular calculations for one highlight, for example). The tradeoff in resources isn't worth the gain for years to come, though... if it were between parallax occlusion or this, what would you choose?
I'd like to see them do it with a less red face. They textured him to look a bit sunburnt, which hides any problems they might have with the translucent look. It falls apart on their adrienne demo. I like the blurred light map, but it needs some color changes based on light being recieved dangit!
Replies
Funny that nVidia is using a technique that AFAIK ATI first used, that is, using RTT and then blurring them to simulate SSS, there is an ATI paper on it/skin rendering.
As for translucency, such as behind the ears, what they are doing doesn't cover that.
This is just a really high-resolution texture (4096 I would guess) and a really high-poly model. I normally love these sorts of demos, but there is nothing new here that a semi-competent shader programmer can't do already, using techniques and samples from shaders that have been out for a while now.
Alright, I just read through the 120-page nVidia paper on this, it is basically the same thing but improved (they mention this in the video). There are some cool things here but nothing new, or groundbreaking, or outside of the box... they're just doing what has already been done but was too expensive to do (6 specular calculations for one highlight, for example). The tradeoff in resources isn't worth the gain for years to come, though... if it were between parallax occlusion or this, what would you choose?
It's still impressive even if it isn't anything new.
if it were between parallax occlusion or this, what would you choose?
[/ QUOTE ]
It doesn't matter cause either one would get ditched during optimization.
We must prepare money for a new computer that he impersonated a decent amount of frames per second in this demo.