Page 1 of 1
Normal map implementation
Posted: 04 May 2010, 00:22
by macuser
kk, was just wondering when normal mapping would be implemented. Olrox told me that one of the devs told him that they would be implemented when there were models that used them. There are plenty now, in fact all of jorzi's models for AR have been normal mapped. Also just wanted to start a topic where the development could be discussed.
-regards macuser
Re: Normal map implementation
Posted: 04 May 2010, 16:31
by Per
That means it gets kicked up on the priority list. Unfortunately, it will not mean it will take highest priority, as most of us think are busy merging in the Qt port, and then new net sync code, or fixing movement problems.
Re: Normal map implementation
Posted: 04 May 2010, 16:41
by Jorzi
Normal mapping would certainly be nice, although I texture my models so that they will work nicely without it as well. I like the challenge of putting as much detail and realism as possible into an unshaded model restricted to ~100 polys. It makes you think of how all those computer resouces are actually used.
I am, however, very excited about technology and I look forward to having my models dynamically shaded.
Some suggestions for cool shader effects that may or may not be possible to implement into warzone:
-Screen space shaders and postpro, this is epecially cool for a strategy game since they do not depend on the polycount or the amount of action, only on the screen resolution. Starcraft 2 makes heavy use of these, one of the most popular being screen-space ambient occlusion (SSAO)
-Specularity and specular maps, this combined with normal mapping can result in some extremely realistic shading. Specularity is essentially an extremely cheap way of simulating the diffuse reflection of highlights from lightsources on materials that we may not think of as reflective, especially plastics.
-Normal vector as texture coordinates, ramp shading. This can be used as a way to simulate the fresnel effect, or dramatic edge highlights. It can also be affected by specular maps to give a nice sense of roughness/shinyness to the materials
Also, if anyone decides to implement normal mapping, I suggest that we use object space normal mapping instead of tangent space, since they work very well with non-deforming models, they don't require smoothing groups and they are computationally cheaper.