Hi, all -
I've been messing around with using true displacement vs. bump-mapping on trees, shrubs, etc. I'm starting to get results I'm happy with and thought I'd show off these test images. Default terrain, etc., just a shot of an English oak. Compare the profile edges of the trunk in the foreground. (The one in the background is bump-mapped in both images.) It's subtle, but your eye notices the lack of smooth edges - really adds a touch of photo-realism. Or I think so, anyway. (I think I saw a post using this technique years ago, but this was my first shot at it.)
The trick is splitting the foliage away from the wood you want to displace. It seems opacity-mapped leaves don't like being rendered with displacement mapping enabled on the object. Done properly, both the wood object and the leaves object have exactly the same origin point, so you put the wood where you want it, copy it's coordinates into the leaves object, and they'll align properly.
So check out the images. I'll be posting a set of models here as freebies in a while - kind of busy at the moment, but I wanted to show these off. I'm pretty happy with the effect.
Nice experiment. You see the displacement especially where the trunk hits the sky. Problem with real displacement is that you need a lot of vertices, which is only worth doing for hero trees and such.
I don't think it adds vertices per se, but the micro-poly renderer may. This was implemented with displacement maps.
I mean that the imported object's trunk needs to be built with more vertices than trunks for bumpmapping. TG doesn't add vertices. The map relocates the vertices for the renderer, so if you don't have enough, it won't do that.
But I believe you're wrong! You don't have to add more vertices. I've attached a shot of the wireframe at the same camera position. That's the whole point of displacement maps - you get the fine detail without adding any vertices or making the model any more complex. Compare it to the earlier versions to really see the difference - both the bump-mapped and displaced versions use the same number of vertices - the micro-poly rendering engine provides the additional data from the displacement map.
To my surprise you're actually right, even with a pretty brutal PF—if set to forced displacement. Apologies for that. I hoped it would work for a population of spheres too, but that is crashing TG.
Yeah, it's pretty outstanding, isn't it? I haven't tried it on populations, so I don't know, but you don't really need to. If your eye sees something with that detail in the foreground (even if you're not really conscious of it) and you see the same tree in the middle distance, your mind just assumes that it's the same. If you study it, an experienced artist will start to pick out the CG effects, but at first glance, you get that instant or two of doubt, and to me that's the real goal in CG art - a piece that'll make you wonder, if only for an instant.