GeForceFX and displacement mapping?
- Thread starter tb
- Start date
Tokelil said:
Actually, I shall probably ask something more along the lines of how HOS is handled on NV30 - does it have a 'dumb' tesselation unit before the vertex array, or can the vertex array create verticies itself for HOS (like P10).
I just realized that one (IMO important) opint hasn't been made here about the order of tesselation and VS.
The whole idea about mixing adaptive tesselation and DM is that the DM should be sampled at each of the output vertices from the tesselation. And I think DM should be done in VS.
Ergo: tesselate first, then VS.
Laa-Yosh:
I understand that using muscel bones and hand tweeking texture shifts would be a pain. But maybe a muscel simulation system like the one you linked to could be used as input format, and then analyzed and "compiled" to a muscel bone/texture shift map/DM - combination that can be done completely at the GPU.
An old idea I had was to use "key frame DMs" for muscels/facial expressions, interpolate between those with pixel shaders, and then use the result as a DM. So you could have a set of keyframes for mouths, eyes, noses, and so on, and blend them together. The key frame DMs could be the output of some HiQ simulation.
What do you think about that idea? (Old news/useless?)
The whole idea about mixing adaptive tesselation and DM is that the DM should be sampled at each of the output vertices from the tesselation. And I think DM should be done in VS.
Ergo: tesselate first, then VS.
Laa-Yosh:
I understand that using muscel bones and hand tweeking texture shifts would be a pain. But maybe a muscel simulation system like the one you linked to could be used as input format, and then analyzed and "compiled" to a muscel bone/texture shift map/DM - combination that can be done completely at the GPU.
An old idea I had was to use "key frame DMs" for muscels/facial expressions, interpolate between those with pixel shaders, and then use the result as a DM. So you could have a set of keyframes for mouths, eyes, noses, and so on, and blend them together. The key frame DMs could be the output of some HiQ simulation.
What do you think about that idea? (Old news/useless?)
