SubD with accuracy in mind – some Youtube resources


Hi All,
there was quite a bit of talk about SubD-Modeling again. One could spot a lot of interest, great faith in McNeel but – at least in my perception – also quite bit of unrealistic user expectation in terms of accuracy one may end up with (or RMA will provide).

I thought some of you might be interested in videos done by German Industrial Designer Claas Kühnen, who for quite a while already uses Blender together with Fusion 360 for Design ideation. He records his sessions for his class at a US-university and is generous enough to share them on Youtube.

All in all he has countless hours of learning material online, including a playlist which starts with the installation of Blender and getting the feet wet with SubD modeling. The linked videos assume some pre-existing knowledge and show mostly unrehearsed and unscripted Design ideation: Freely playing around with forms, without any sketches, directy in 3D – and if possible in ways which leave a lot of room for later changes.

Modo or Blender - from a Rhino users perspective?
(Nathan 'jesterKing' Letwory) #2

Nice to see some Blender mesh modelling with the subd modifier in use.

I just wanted to clarify that where Claas is saying he is sculpting he is basically just mesh modelling with the mesh being the control cage. Sculpting in Blender world is a different thing (but also mesh modelling, just not like in these videos).

Anyway, these mesh editing tools and workflow is what I definitely would be happy with in Rhino, also on the SubD modelling front :slight_smile:



Yeah, he is not sculping in the sense that he deforms a far denser mesh with brushes – but then
again also Nurbs modelers sometimes call controlpoint editing of surfaces sculpting :o)


@hifred Sculpting or better CV sculpting is a term that is commonly used also along Alias users for perfecting Class-A surfaces where you need to align the surface patches via individually manipulating the CV layout till the surface continuity is perfect.

So in blender I sculpt the CVs and in T-Splines I still sculpt the CVs - construct patches either following brute NURBS limitation quad patches with even topologies for smooth sub-d to NURBS transition or irregular topology focusing more on what Sub-D modeling is like.

There are a lot of things you can do in a system like Maya or Blender where Modo or Rhino simply pale which is why I pair Blender together with Fusion 360 where it offers true and full parametric modeling and

heheheh fillets that work :slight_smile:
(former Rhino user)

If you want sub-d like work in Rhino there are many options already you can choose from and if you really want to embrace it Rhino needs to get some design history for both NURBS and the SUB-D module.

I kinda use Blender as a generative modeling tools similar to Grasshopper but more like Fusion.


Depends…maybe yes, hopefully not. The goal was:

  1. Highlight a well known need, gauge interest.
  2. Encourage developer focus on innovative interface design to advance accurate sub-d creation as competitive advantage.

Yes, interesting links. Thanks!


what do you mean when using the word ‘accuracy’ in this context?

because for me, if i’m talking about ‘accuracy’ in mesh vs nurbs, i’m talking about what’s shown in this simple example:

the blue curve is an arc @ 112"… the red line needs to intersect with the arc and it does so at a length of 107.25"

the same thing only using polygon style:

so, you see? there’s over 1/2" inaccuracy in the meshed model… and this is just a simple example… once you throw surfaces into the mix and/or more complex shapes, it gets even worse… to the point to where you simply can not trust anything about your model if attempting to build…

i feel like you’re meaning something different when saying accuracy because, to me, those example videos have very little to do with being accurate… i don’t think they mentioned a dimension or length at all.


@jeff_hammond Your flaw is that you compare NURBS to polygons. Polygons will be the low cage you can edit the CVs the same way as with NURBS by the way. Inside the Low Poly Cage a smooth surface can and will be skinned with NURBS and T-Splines.

The main difference if you are an Alias user is how clean you want your NURBS patches to be. That is the down side of TS and all other SUB-D to NURBS systems. The individuals created NURBS patches are pretty heavy (spans).

Otherwise TS created BREPS are precise as they can get. They are NURBS afterwards.
I would build an axel with TS so would I not use a CV circle for that as well.

However Class A surfacing for cars is required but a painfully slow process. If you are good at sculpting meaning you are really good at modeling and have trained your eye you can build fluid shapes in a fraction of the time and produce G2 smooth surface patch layout.

It is a compromise one has to do. I think in general engineers hate TS because they want to rebuild everything. We brought many products on the market where TS was used without any issue on the surfaces.

Obviously you need to know how to sculpt good to maintain proper topology.

Just to day I sculpted a housing for a rat and air flows in minutes which would have taken me ages in a parametric sketch modeler.

I would really see something like sub-D driven NURBS patch creating as the missing link between NURBS and solid modeling tools because those two also have drastic limitations.

Why is it that so many say sketching is faster? Simply because they are not good at digital design tools and use also the wrong tools for rapid ideation.

Ah wait they use SketchUp then hahaha.

The car industry embraces Sub-d crazy. I got 4 of my students into Ford here in Detroit doing concept modeling just with what I taught them in Blender because the concept of NURBS and SUBD are the same to all apps.


flaw how? the example is flawed?

why is it a flaw for me to compare nurbs to polygons when the OP, hifred, is (i think) talking about polygon modeling?

idk, someone here is misunderstanding something (probably me) because you’re counter arguing what i’ve said with t-splines… i could fully use tSpline surfaces… but it seems to me that you think i’m saying surfaces created with t-splines are inaccurate and unusable.

(Nathan 'jesterKing' Letwory) #9

I think it would make sense to keep in here talking about the control points (the cage), not the way a program generates the surfaces out of those. Obviously NURBS will be much more accurate, but I’d say comparing the control point editing would be useful in finding a good workflow doing SubD-modelling.

The original post doesn’t really talk about polygon modeling, although Blender pretty much is a polygon modeling tool. The point is that SubD modifier is used, so you essentially get implied surfaces. Here you do SubD modeling, just with access to really good mesh editing tools - since the cage is just a mesh really.

To export a good cage I’d use the base mesh that in one manipulates in Blender, not the resulting mesh. It is the cage that matters. And then you can get the accuracy you need in Rhino after iterating through your design in Blender.


but there’s more context if seeing the prior posts in other threads which led up to this one… for instance:

start here and read the 25 posts or so that follow… there is a clear distinction being made by hifred regarding mesh & nurbs surfaces…

Robert McNeel & Associates are in decline
Robert McNeel & Associates are in decline

i used david’s post as a launching point since it’s the one that brought the word ‘accuracy’ into the mix… i feel hifred isn’t understanding the word in the sense being discussed which is why i put an example of ‘accuracy’ in this thread


Well, I think the thing is here we’re not looking to do away with precise, mathematically derived geometry like lines, planes, circles, cylinders, cones, etc. Mathematically defined stuff will continue to exist because there is a huge amount of stuff out there that does not need any kind of free-form input, and/or needs to be manufactured to very precise tolerances.

We’re talking here (I hope) about adding an additional system that facilitates “sculptural design” (for lack of a better term) while remaining somehow compatible to the existing “mechanical” ones in Rhino. Like being to able to design some sort of swoopy surface for an injection molded part with SubD or whatever and still be able to punch perfectly cylindrical H7 dowel pin holes in it.

The user will have to beware of what tools they use and their inherent precision, but that’s already the case now with NURBS free-form surface creation/editing.



Correct Jeff, the word ‘accuracy’ has been assigned to two distinct but overlapping aspects - accurate conversion and accurate design intent.

The former is a given to the extent possible by the underlying math and comp sci; the latter, while related to the former as well, may possibly be derived somewhat through creative application of the UI design.

At least that’s how I see it from the other side of the fence. I don’t write the code.

edited for mobile device typos


actually even your drawing which shows the mesh is too accurate :o)

On the mesh one only knows the coordinates of the 9 points which make up your blue section line.
Don’t worry, I’m not mixing up things here.


Thinking something similar…

My view: an advance to sub-d does not necessarily mean accuracy to .00 whatever, rather, design accurate forms, capturing design intent, in a manner which approximates design accuracy of NURBS, as an advance to sub-d to the extent possible.

For example, designer knows a region of a form should have a radius of 10". Sub-d interface will allow designer to input 10". If the resulting NURBS form is 9.98" at input region, no worries, as that is ‘design accurate’ sub-d.

There are many more examples. For inspiration, all one needs to do is think about what design intent is possible with NURBS, then think about what general process approach (not tools) might potentially be applicable to sub-d design intent. The easier and faster one may create design accuracy…bingo…advance!


A few words on accuracy (mostly as an answer to Claas)…
There’s no doubt that one can more or less directly manufacture a variety of shapes which are converted to Nurbs from mesh cages.

In work areas, such as ornamental jewelry one can even skip the Nurbs conversion altogether and use a HiRes mesh for series production. This jewelry Designer IIRC started his 3D modeling career with Rhino and now only uses Zbrush to create his complex shapes. As the next step after modeling is wax printing one here can also safely skip “properly” assembling things. For as long as the printer finds a valid contour, it will print out just fine. I’m frankly quite puzzled that so few jewelers use mesh modeling programs…

In case one does convert to Nurbs from mesh cages one usually will have the greatest success with parts which only require a fairly limited amount of surfacing operations downstream. A swoopy seat shell which needs some exactly round holes or a clamp to mount it to a metal frame – things of that nature. What’s cool in Fusion’s SubD implementation is that the sculpted geometry is treated as a Feature and may get changed even after having adding some solid modeling operations on top.

Making a series product from your hair dryer concept sure was a far greater challenge. If one needs parts which sit cleanly inside each other, have joints, hinges, threads, ribs and domes and all that stuff one will rather sooner than later pay the price for the complex and unorthodox surface layout of auto-converted geometry. In many instances this will realistically equal complete remodeling – but one at least has already a great concept model to begin with.



As long as designer can get his/her swoopy contours to within perhaps .06" of their intended curves, and within perhaps .06" of their intended locations. The rest can 'fill in" as with Network Surf.


He does not talk about using a polygon ring to represent a nurbs ring.

This is about - sub-d like surfacing and creating NURBS geometry from it
converting the smooth SUBD to NURBS not a linear surface to BREP.

While actually I also used just polygons for furniture design instead of anything
else because it vastly out performed Rhino or Fusion doing the work where polygons
were just dramatically faster to work with.


That is why I pointed out proper topology.

If you really are good at class a surfacing you know that there you face the same challenge.
Also with NRUBS you can build a design later the engineer even cannot use.

T-Splines in addition also offers help to layout the patches better if desired.

I taught Alias for years and used it for work and honestly while it out performs Rhino on multiple
areas (precision and such) the process itself is a nightmare.

Would I build a car with TS obviously not. Consumer products? Can easily build with using TS and
the other surfacing tools and after production you will not see what was used.

The trick is not to use something because it is easier but to use it right!

TS NURBS SOLID they all have ups and “downs”

How the industry will adapt is in addition a completely different questions!


There is also another thing to point out why I sometimes see people reject such ideas.

While TS geometry can have issues “if you use it the wrong way” I see the main reason why
people not getting it or rejecting it because they honestly don’t know much about it or are in
general pretty poor at CAD using it as a design tool and being fluid at surfacing as they are
with sketches.

I am with TS since it came out in Maya then Rhino and now is part of Fusion. NX Catia ProE
they all have similar modules build in. Some only use it to explore maybe an idea. Other use
it to simply generate with TS the surface that is too hard with NURBS and profiles and then
create other tools with surface or solid modeling tools and workflows.

In my teaching I strictly tell the student that I do not teach them an application but tools
and a workflow to use them right. So we mix Alias surfacing, sub-d sculpting, and sketch based
parametric modeling were is just also makes sense

Instead if focusing that everything has to be rebuild and constraint afterwards which I think is
often a problem as well.

Obviously from project to project this can be different. Designing a car, plane or a toothbrush or
playstation controller are all projects with different complexities and requirements.

We just found that TS never gave us any trouble when bringing the products onto the market.


What I find interesting for Software Makers though…

You (and other advanced SubD-modelers) still greatly prefer creating their control cages in 3rd party apps,
although Autodesk built Fusion pretty much around Tsplines. You use this module mostly as a converter.

That also means that you are not actually using Tsplines btw.
It’s still the product name yes – but you* seem not to be using the Tsplines local subdivision scheme which
was subject of several papers and the starting point for the whole Tsplines project…

Just some observations.

*neither do I