T-Splines: The new way of modeling


T-splines are a revolutionary patent-pending surface type. NURBS and subdivision surfaces are both special cases of T-splines. Thus, T-splines inherit all of the respective strengths of NURBS and subdvision surfaces (SubDs) while eliminating most of their weaknesses.

A T-spline can be thought of as a NURBS surface for which a row of control points is allowed to terminate, without traversing the entire surface.This simple idea imbues T-splines with several significant advantages over NURBS, such as the ability to eliminate superfluous control points, to perform true local refinement, to merge two NURBS surfaces, and to add local features

A T-spline can be thought of as a NURBS surface for which a row of control points is allowed to terminate, without traversing the entire surface. This figure (left) shows a T-spline control grid (or T-mesh) next to a NURBS control grid. This simple idea imbues T-splines with several significant advantages over NURBS, such as the ability to eliminate superfluous control points, to perform true local refinement, to merge two NURBS surfaces, and to add local features.

Cubic NURBS and SUBDs can be converted to T-splines (or T-SubDs) without any error. Likewise, T-splines can be converted into cubic NURBS without error. This means that T-splines integrate effortlessly into modeling environments that involve special-purpose NURBS or SUBD programs.


Sounds interesting. It certainly would be nice to have just one modelling method instead of having to mix NURBs, meshes etc. But I’m not convinced it will work because it’s still too much like NURBs and NURBs control points are usually much harder to work with IMO than a subd cage.

Also, I see no mention of texturing. NURBs are well known to be less flexible for texturing than subds. But I guess that a T-spline doesn’t use a full rectangular control mesh means UV projection might be necessary.

For now, I’m quite happy with subdivision surfaces because I can model low poly meshes to use in a game engine as well as have subd versions of the same model for pre-rendered animations. It doesn’t look like T-splines can do that - it seems geared solely for pre-rendered graphics.

It’s early days yet though.

Hrm, patent pending? That’ll insure it doesnt go anywheres unless in a major package. And that’ll be fun for open source, heh… </sarcasm>

looks a bit similare to what i used in hash’s AM yeasr ago. there you could add splines into a patch to make it more dense.


yawn. to many points still… The real new flow is Zbrush and silo3d at the end. Zbrush to make a massive mesh of details and silo’s topoligy brush to make a super clean mesh from that mess.

patent-pending? A particular implementation maybe, but the general idea can’t be patented, and more generally I am not sure you can patent ideas, at all (?).

This particular idea has been done too many times in too many ways for anyone to patent it anyway.

Your not from the US are you?

Seems as of late, anything can be patented, real or not, prior art or not. Just look at all the patents Microsoft files for despite massive prior art.

Recently they filed a patent for Virtual Desktops, something that SUN came up with a couple of decades ago.

Not long ago, someone actually won a patent for inventing the wheel! It was just phrased in such a way that it got past the patent lawyers. Then theres that one company that was trying to file patents for segments of the human genome they had discovered. Ugh, the list goes on.

Theres a lot of patents on ideas too. The marching cubes algorithm for instance (which is used in 3d) . Theres a number of patented algorithms that have made programming very difficult in the US.

Luckily for us, it doesnt apply a whole lot to Blender as it can be said that Blender doesnt have to respect said patents as its not of US origin. However more and more countries (like Australia and a lot of Europe) have been signing agreements to respect said patents.

IT use to be patents were to protect an inventors ideas for a limited time. Now they’re used by multi billion dollar companies to eliminate competition and have monopolies. Its particially effective against open source efforts, given most open source projects lack the funds to go to court.

It’s just decimated control meshes for NURBS, big flipping woop.

The patent system in america is fs checked.

well if the US would adhear to the PCT treaty a bit better, then things would be a bit easier IMO.

its an international patent group which you can go through to enable a patent in all countries. i would rather this system than a monolithic international patent (i.e. you get a patent in the us and i applys all throughout the world)

however i can use some of the US patent system to my advantage. :smiley: there are a few rules i will be playing.


Hmmm…Isn’t this patches, after all? Can be till 5 points, I think.

That is, Hash:Am , the japanesse Hamapatch (which I’ve used) , Spatch(also used it), Jpatch, all those modellers use that already old technology… Even Max, and Metasequoia, have a modelling mode with patches, though these two last i dunno if they’re actually real patches.

Patch modelling have a serious problem with creases, at least till you know some tricks…

I kind of prefer subdivs… mainly work in the low pol cage most time…

ExtrudeFace: is your signature still true? I can’t believe that.
Patent Smatent! How will companies stop random Cvs patches?

But implementing T-splines in blender?
Hmm,… what about normal splines first? Trim, intersection and other boolean functions still seem a hard coockie to solve for blender developers.

The Westerner game, recent one, was fully made using patches. It’s even possible in rt 3d games. I think q3 levels has some sort of support of it.

In packages I know, Hamapatch, etc, I find the modelling a bit limited compared with polygons, and subdiv. But it may be just me, and anyway it depends on how is implemented. Many stuff not useful in other packages, is useful in Blender for how it’s been made…[and in some cases, viceversa…:wink: ]

"ExtrudeFace: is your signature still true? I can’t believe that. "

Thanks , that sounds as a cumpliment :slight_smile:

I’m a total newbie in the sense I am also in most packages : I only learn what I go needing each moment for each package…“survival” style :wink:

There’s allways lots of stuff that I never learn of a package because a job, or a home project, do never require it… But in the other hand, I try to get allways to the maximum in those things I actually need to get quality in a project. For example, I have worked a lot with max, and I only know the edit poly, edit mesh, skin modifier,material editor, character studio and a bit of other stuff…no idea on post processing , professional rendering, particles, etc.

In Blender, I have only learnt to deal with basic character animation, and only know the 2% and what I need to export into real time 3d game engines. Oh, and of course, UI in general, and some modelling (curiosity. To check that at the end, I could us it for modelling too if feel in the mood)

I explain as I have often thought of changing the signature, but the fact, it seems it will keep being allways true.

The only thing I could say is by no means I’m newbie in 2d/3d, in general, but that’s all :wink:

This looks nice, and will probably replace nurbs eventually, but I have reservations as to whether or not t-surfs will replace sub-divided surfaces.

The nice feature is what the website spends the most time talking about - adding localized detail to a nurbs model. So t-surfs seem to have all the goodness of nurbs plus removing a major frustration.

The part I have reservations about is the feature the website tends to gloss over - the ability of subsurfs to have an arbitrary topology. All the isoparms of these t-surfaces are laid out on a rectangular grid. True, an isoparm doesn’t have to go all the way from one edge of the patch to the other, but the basic layout is still rectangular. This doesn’t allow me to, say, merge two control points together, the way I can with subsurfs. Also, when I’m modeling with sub-surfaces I often take a polygon and extrude & scale it (such as when modeling the eye area of a face) - this does not seem possible with t-surfs.

Of course, I may be wrong (at the bottom of this page it says sub-surf implementation is not complete). If so, t-surfs could become quite nice.