adaptive tessellation workflow

DrillerKiller

Does anybody have any experience or pointers for the art side work flow with adaptive tessellation? I have a basic understanding of it, but from what I've read it can be tricky. Any input from you guys would be greatly appreciated!

DrillerKiller

come on, nobody?

r_fletch_r

perhaps you should explain more about what you want. Do you mean sculptris?

SpeCter

r_fletch_r: i think he wants to know know if there are some special workflows regarding the "new" tesselation features introduced in DX11.

DrillerKiller

you bet spec

Surfa

Tesselate modifier in 3ds max plus displacement mapping?

SpeCter

Someone correct me if i´m wrong, but i think it´s not something you really care about when modelling, the tesselation is completely shader based.
The algorithm uses different techniques for different cases, like planar approximation between the high poly and a plane(how much difference in height is allowed etc., like you are near the model, less difference is allowed resulting more detail and the other way around of course).

I think other things the algorithm takes into place is the maxsize of tesselated polygoins´s (smaller ones = more detail)
and the other one is angle based.Let´s say you have a brick wall depending on the maximum angle which is allowed more depth in the bricks is allowed.

Hope this helped a little bit and isn´t full of shit

Ark

Think it just uses a traditional displacement map.
Theres a whitepaper around about Unigine's implementation that shows their material settings.

PhilipK

Haven't worked first hand with any of it so far, but as far as workflow goes you should, just like when you're preparing a mesh for a sculpting prog, try and have an even tesselation on your "control mesh", since this will be subdivided. If you take a sculpted piece as an example you do things as usual: diffuse, spec, normal. But you also need a kind of tesselation "control map" which contains info you bake out from your highpoly as well. Other than that you still got limitations like trying to stay out of extreme angles when comparing low to highpoly.

Remember this may not really be accurate, this is just how I understood it to work in general! I also bet there are several differences between different companies using this.

Also I think we'll see a lot of games doing this the same way parallax stuff was (ab)used in the early on, like just adding extreme offsets to stuff like a regular brick wall causing super ugly stretching on 90 degree extrudes etc. But I'm really looking forward to see more stuff utilizing this stuff.

Eric Chadwick

Check out the DirectX11 presentation at the bottom here.
http://www.polycount.com/2010/07/01/graphics-from-the-future/

Some visual examples in there of different tesselation routines. You'll see why I dislike PN Triangles... looks crappy compared to a subdivision surface. Might have a benefit though, I bet UVs aren't so badly messed up as with sub-d.

DrillerKiller

thanks guys

r_fletch_r

Question for someone who understands the tech: will this make it possable to do Catmull Clark SDS on the video card?. the idea of finally using sub D in games is pants wettingly fantastic. That and all the ram/fps it could save in 3D apps displaying SDS previews.

[edit] The man in the presentation says its possible .:)
I must say PN Triangle looks like shit

Eric Chadwick

I wonder if they get around the issue of subdivided UVs creating gaps/seams? I imagine you would want the UVs to be subdivided along with the mesh, but that means they're going to pull away from their initial corners.

r_fletch_r

Eric its really easy to fix this problem before it happens. UVlayout has an option which transparently compensates for the stretching. You just unwrap as you would normally and the final layout works a charm.

I dont know expensive it is to do during SUBD but I believe blender preserves the UVs on SubD.

Michael Knubben

Pn Triangles is just the same old shit they've been trying to sell for a long time now. If the content isn't made to be tesselated, you're not going to get a better result than the lower detail version that was specifically authored by an artist.
Where Tesselation gets interesting is with the use of heightmaps or sub-d! Sub-d would be useful for characters especially, given the limited amount of vertices that would need to be skinned, although ofcourse sub-d for organics has largely fallen out of favour, being replaced by sculpting software.

r_fletch_r

You still need to retopo so the sub d stage wont be such a big step.

A couple of cool things stick out to me about this tech.
Terrain is going to be a million times better. If Vector displacement is used then overhangs and all sorts of cool stuff will be possible.

Also I wonder how far people will be able to push a pretty generic mesh, relying on displacement for the details. I wonder will it get to a stage where you see very individual looking crowd characters all using the same mesh but different displacement maps/blends.

mortalhuman

ah, interesting indeed fletch.

DrillerKiller

interesting stuff guys, for sure.
I am skeptical of the whole process, PN is shit and imo only really useful in a hand full of scenarios, sub-d sounds good but opens up another can of worms for an artist pipeline.

jocose

Yeah I either A) want to get my hands on it and play around, or see some more in game final asset examples.

Until then its hard to quantify its usefulness in terms of strengths and weaknesses.