requested feature: Subdivision to Nurbs

[c-g]

Maybe you should have asked for Modo to actually have NURBS first. :)

[paul k.]

why nurbs?


Pretty much a dead system< No?

[jamesvalue]

I think that T-Spline are the way to go for future modelling , and of course a R. engine built inside Modo , capable to render every producted geometry .

Bye

[jamesvalue]

You can go here to get more infos about t-spline:

http://www.tsplines.com/

Bye Luka

[sacslacker]

Sub-Ds or poly modeling seem to work for everything I need to do. I personally don't like NURBS.

No clue on T-splines really but I'm betting it's going to take a bit of work to push Sub-Ds and poly modeling out.

[jamesvalue]

They sayed (T-spline web-page) Subds and NURBS are all incleded inside T-spline modelling!


Is that believable or not?

Bye Luka

[c-g]

T-splines are poised to become an industry standard.

According to?.....Them?

[paul k.]

At first glance I don't like the idea of T splines. I find that when I really model well I follow the sub- D rule of getting as much as you can out of the least amount of geometry and then refine for more. When I overdue it, it really comes back to bite me. Instead of modeling high and then reducing with T splines why not just model well and not have to use them at all?

I guess I could be missing the boat herre if they do something cool but from what I saw I just a saw poly-control reduction as a opposed to your standart poly reduction tools.

I mean if it helps people model better then I'm all for it but I don't think at first glance it will really help me all that often.

[StephanD]

T-Splines looks like a nice concept and I understand it is still Beta but it's far from ever becoming an industry standard...Why?

Because first,the optimized mesh is still heavy and contains too much N-Gons.

But that's not really the point.

Fact is,you could do this on polygons using a clever curve/peak angle detection algorithm that will identify and remove unnecessary edges or polygons.Don't ask me how it would be done though but I already do it manually.

Good idea though.

[nurcc]

Hi there.. I'm a guy who works for TSplines. In fact, it looks like I'll be one of the guys at the siggraph booth the year, to answer technical questions.

Maybe I can answer some of your questions. The first is about the phrase 'poised to become an industry standard'. Yeah, we hope that'll happen, but we realize that it's not going to be next week or whatever. Maybe in 10 years, if things go well. But we certainly think that T-Splines have advantages that can be exploited now, and we're trying to get them out the door asap.

About the reduction - that was a paper that we used to show that converting from Nurbs to TSplines can save a lot of space. However, we imagine that the workflow more often be to set up the basic topology, and then add detail in T-Splines.

About Nurbs and Subds - there's a little explanation here (http://www.tsplines.com/nurbs_and_subds.html) and more here (http://www.tsplines.com/FAQ.html). The basic idea is that Nurbs and Subds are related, but incompatible. With T-Splines, we combine the two - not the basic ideas, but the actual math and algorithms. For example, if you convert a Nurbs to a T-Spline, you get the same CVs, and the exact same surface. If you convert a Subd to a T-Spline, you get the save verts, and the exact same surface. On a T-Spline, you can do things like extrude, as well as inserting partial isoparms that don't change your surface. We're pretty confident that'd be useful for CG as well as CAD/CAM, it's a matter of making a good implementation and tools.

About ngons and the mesh examples, the meshes were converted from Nurbs, and as such don't have the nicest topology. Of course, you can get lighter meshes with better topology. One thing about ngons is that those faces shown with 5 (or more) points aren't what you'd consider ngons in SubD modelling. They don't pucker, shade weird, or behave oddly in other ways - they just act like quads. It's probably hard to tell how they behave without getting your hands on them.

Finally, I have a couple of pictures to show. These are T-Splines that an artist made for us. The first one is an example of SubD topology. The second is a similar model, but with T-Splines. In both models, you've got a denser mesh near the fingers, and it's lighter near the wrist, which makes sense - less detail in the wrist. In the SubD topology, he avoids ngons, and uses poles instead. In the T-Spline, we ends the edge loops by just.. ending them in Ts.

http://www.tsplines.com/products/images/Hand_Tut_1.jpg

http://www.tsplines.com/products/images/Hand_Tut_6.jpg

Hopefully this helps answer some questions..

Also, I realize that this is a modo forum, and this is a bit off-topic. Sorry if it's annoying..

[jamesvalue]

Does a "T-spline based geometry" rendering engine exists?

Bye

[nurcc]

Does a "T-spline based geometry" rendering engine exists?
For rendering, you can either tessellate to a polymesh, or split into several Nurbs or Subds. We also have a Renderman output option, which outputs as bezier patches. So, it should work with pretty much any renderer out there.

[paul k.]

The basic idea is that Nurbs and Subds are related, but incompatible. With T-Splines, we combine the two




But as I was saying earlier- Aren't Nurbs Dead?

I heard WETA was the last to use them on a large scale and even they switched to Sub D's...

I mean I'm sure their's still someone, using them but overall it's pretty much a dead method. I was even at an XSI demo like two years ago and the guy was showing the sub d capabilities in XSI and as he was doing it he was like " I call this demo...... Nurbs are Dead ". Everyone aplauded at the idea that we wouldn't ever have to deal with them again.

I guess it comes down to whatever tool works well for you, but for speed, flexiblity, and overall efficiency, I think Sub D's have an overwelming case. http://www.cgtalk.com/images/icons/icon7.gif

[jamesvalue]

For rendering, you can either tessellate to a polymesh, or split into several Nurbs or Subds. We also have a Renderman output option, which outputs as bezier patches. So, it should work with pretty much any renderer out there.


Interesting!

I always thought that a rendering engine could strictly only renders polygonal surfaces even in case of NURBS or SUB-ds meshes (I.E. , Maya/Mental ray or Rhino/Flamingo) , but does Renderman renders bezier patches also?

(sorry for my English)

So something somewere start to move!

Bye

[Sil3]

But as I was saying earlier- Aren't Nurbs Dead?



For high precision work they are still used.

[nurcc]

But as I was saying earlier- Aren't Nurbs Dead?

I heard WETA was the last to use them on a large scale and even they switched to Sub D's...

I mean I'm sure their's still someone, using them but overall it's pretty much a dead method. I was even at an XSI demo like two years ago and the guy was showing the sub d capabilities in XSI and as he was doing it he was like " I call this demo...... Nurbs are Dead ". Everyone aplauded at the idea that we wouldn't ever have to deal with them again.

I guess it comes down to whatever tool works well for you, but for speed, flexiblity, and overall efficiency, I think Sub D's have an overwelming case. http://www.cgtalk.com/images/icons/icon7.gif

I'd agree with you that Nurbs are pretty much dead in most of the CG world. There are still a few CG shops that use them almost exclusively (like Dreamworks) and a lot of places use them for props and mechanical modelling, but not for organic shapes. They're also pretty much the only thing used in CAD/CAM applications, which is a much larger industry than CG. If you're looking at going from a nice modeller like modo to actually manufacture something, you'll need to convert it to Nurbs, like the original poster requested.

I think it's pretty easy to understand why Nurbs died - it's the arbitrary topology of SubDs. You don't have to insert whole lines, you don't have to make patch models to get the topology you want, and the resulting surface is really nice. The major downside of SubDs used to be that they didn't integrate with the rest of the toolchain, which has eventually been solved for all but CAD/CAM, and that there are some operations that you could do with Nurbs but not with SubDs - for example, inserting an isoparm. It's possible to insert an isoparm into a Nurbs without changing the surface, but if you split a bunch of faces in a SubD, it'll pull the surface up towards the cage.

Rather than thinking of T-Splines as Nurbs with arbitrary topology, it's probably better to think of them as SubDs, where you can add geometry without changing the surface, and avoid ngons and poles more easily. They're also more compatible with Nurbs, in that you can take any Nurbs patch model and merge it exactly into a single T-Spline, without any approximation, which is impossible to do with a SubD.

In the end, it's really about the technology that meets your needs and helps you get the job done better/more quickly/cheaper. Hopefully, we can get T-Splines to the point that they improve on SubD modelling as well as Nurbs modelling. And while we think the technology is promising, the only way to tell if it's actually useful is to get your hands on it and see if it helps.. :)

[nurcc]

I always thought that a rendering engine could strictly only renders polygonal surfaces even in case of NURBS or SUB-ds meshes (I.E. , Maya/Mental ray or Rhino/Flamingo) , but does Renderman renders bezier patches also?

Renderman is interesting in that its rendering primitive is a micropolygon, which can be thought of as a bunch of points sampled from smooth surfaces. They started out with support for patches, Bezier and other kinds. In fact, for a long time, they treated polygons as flat patches, which is why polygons were actually slower than the equivalent Nurbs model to render. I think they've sped things up in recent versions.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.