yeah, you lost me...but i still read it...and i loved Shrek, so that's all i have to add...;)

Lilly I am just talking about real-time inverse-kinematics animation of Nurbs using a deformation skeleton. It is not rocket science ;)

Yeah Jake, I mean can't people realize that multiple animated Free Form Deformations combined with bump-mapped surfaces and a three point lighting system can be fun?

Oh, ofcourse, I know that real-time inverse-kinematics animation of Nurbs using a deformation skeleton comes up in my everyday conversations all the time.

You must hang out with 3d Animators then. ;)

Darn... my sarcasm came back and bit me... ;) I acutaly mostly hang out with band geeks. Though a few of them do some 3d graphics modeling. But I really doubt thats much like animation.

I'll have to throw that into some conversation sometime and see the outcome then wont I? :D

well, now that you put it that way...it all makes sense...:P (nutjob)

To say that seeing Shrek is to see the effect of NURBS is a bit inacurate. If I'm not suffering a massive hemorage, then Shrek was rendered using Pixar's Renderman (PRman, probably version 8, 9, or 10). The program takes in NURBS, then converts them to subpixel polygons. So, you may use NURBS in the Maya screen shots on the extra features of the DVD (ahh, but are they really NURBS, or are they once again NURBS converted to triangles? Probably triangles), but ultimately you are seeing triangles on the big screen.

OK, really you are seeing photons reflect off the screen, and those photons are effected by the film which was generated at some point by a file of pixels. But, those pixels were (partly) generated from triangles.

So, back to the xbox. Is it possible to do NURBS on it without converting to triangles? Possibly you could do it using programmable shaders, but I wouldn't count on it. The CPU could do all the rendering without using triangles, but that would be slow.

However, by comparison, one can easily do NURBS on the PS2 without resorting to converting them to triangles. This is because it doesn't have a hardware rasterizer, instead it has a chip that is optimized for doing rasterizing, including acceleration for 3D functions. What it takes as inputs to rasterize is whatever you write the software to rasterize. It just is probably too slow to actually use in any games.

A related question would be, why would you want to render NURBS directly using dedicated hardware? For the chip real estate it would take to do that, you could do all sorts of things that would be more usefull to everyone, not just the people who want NURBS.

OK I will have to do more research. I know that most of Shrek was modeled using Nurbs and I'll have to see if I can find some screen shot of those Nurbs being animated and rendered. Luckily I have reader who works at PDI so he can clarify how much Nurb work on Shrek was kept as Nurbs. I just think it is fascinating to see that the custom chips in the Playstation Portable were designed to handle Nurbs and polygon models natively. The use of real time Nurbs in games is still rather academic Siggraph stuff right now but it could be big. They are already using particle effects in games so I don't see why not to use Nurbs. I know the Quake 3 engine was designed to handle curved surfaces and wonder if they were talking about Nurbs.

I believe Q3 used bezier patches which continuously tessellated as the 'camera' approached them to approximate curved surfaces.

Q3 uses curved surfaces. I believe that it specifically uses bezier surfaces. NURBS tend to be viewed as more computationally intense than bezier surfaces. The formulas representing NURBS certainly tended to look less warm and friendly than the ones for representing beziers.

As to the PSP using NURBS, I don't know anything definate, but I'd guess it is just a hardware tesselator. Guess what? OpenGL has built in tesselation features, and any card can support it (though few consumer oriented cards do, unless these new programmable cards add support).

As to Shrek, I'm not arguing at all about how Shrek was modelled. I'm just saying that if they used renderman, like some places claimed, then it was converted by the renderer into triangles. If it was rendered using another program, there is still a very good chance that this is what happened. Traditionally, it is very difficult to render NURBS with a scanline renderer without doing this. I'm not aware of any recent research changing this. However, it is fairly trivial to directly raytrace NURBS. All other things equal, the results are just as good either way if one is tesselating the NURBS to triangles that are smaller than the pixels used.

"Think of how Shrek looks and moves. That is Nurbs in full effect.

Now think of Laura Croft in Tomb Raider 1 She is about as polygonal as you can get."

So you have nurbs in full effect compared to a very low poly character. Shouldn't you give an example of both in full effect.

Heres an example so you can compare the two.

http://www.unrealtechnology.com/screens/character_creation1.jpg
http://www.unrealtechnology.com/flash/technology/ue30.shtml