Future of T-Splines Pixar SubD in Alias

[cekuhnen]

I am curious about T-Splines future in Fusion.

Since it’s introduction T-Splines fell short on delivering a functional hybrid modeling environment. While the main modeling tools are fine every aspect of T-Spline matching a nurbs surface is still barely working well. Also I found snapping t-splines onto scanned meshes very frustrating.

It seems also that the Alias folks just gave up on Speedform because Alias now switches over to Pixar’s own open source SubD System delivering a hybrid modeling experience that is truly terrific - unlike T-Splines BREP in Fusion.

https://forums.autodesk.com/t5/alias-forum/whats-the-difference-between-subdivision-in-alias2020-and...

So any plans to throw out T-Splines and replace it with what Alias uses?

that for sure would increase the modeling value of Fusion which among the sketch engine still is a somewhat weak area in the application.

[TrippyLighting]

@cekuhnen That sounds like a match made in heaven to me. On the one hand we have a team that manages to make the limit surfaces in OpenSubdiv available for true hybrid modeling with NURBS surfaces in Alias in 18 months.

On the other hand we have a sketch engine which still after 7 years of development has not been able to implement stable and reliable spline behavior. OK, granted it works sometimes until they break it again 😉

[Intuos5]

Coincidentally, I just went through some T-spline ideas regarding additional modeling tools. Quite a few of these were added to the backlog or were accepted years ago. Remarkably, none of these were implemented in Fusion, so did development cease on T-splines?

[Intuos5]

I thought I should add this to the discussion from this topic dating back to 2016, which is certainly interesting for anyone regarding t-splines:

---

@PhilProcarioJr wrote:

I would like to point out that it was stated two weeks ago that there is no work being done to T-Spline tools in the near future. Quoted from Colin.

"As for T-Splines, we have nothing in the short term with regards to tools, however we will be working on improving the T-Spline performance."

Given that only recently, I started learning Fusion, you guys know this better than me, but I haven't seen much improvement in the t-spline toolbox since then. I'm actually hoping that anyone from the development team could shed a light on the future of t-splines — and no, not you Chad! 😉  — since despite having used the tools much yet, I can already see lots of room for improvement and fingers crossed there will be.

[PhilProcarioJr]

@Intuos5 

Wish and want all you want to just don't hold your breath or you will find yourself waking up on the floor. 😉

[Intuos5]

@PhilProcarioJr  I feel the frustration you voiced in that topic and I am with you in t-spline requests ;). Luckily I have yet to sit through a couple of years, while still scratching the surface of the new modelling capabilities opened up to me by Fusion in the meantime. I come from an architectural background (Revit) where the modelling options are like this:

and lofts can only be controlled by editing the profiles with no way to control curvature, the borders of the loft are next to uncontrollable, heck they cannot be trimmed precisely. Not to mention that lofting itself is considered 'advanced' stuff'.  Let alone the pride someone expressed in demoing a competing software package by showing a sweep of a curved profile.

Excuse me for the sidestep it holds relevance in a bit. I was thinking about the other programs that have t-splines incorporated. Revit, which I referred to, has no proper dedicated surfacing tools and don't expect them to show up, trust me. Dynamo, which originated as add on to Revit, has received t-spline support in the past year or two (maybe three). So only those who do Dynamo programming and do not resort to Grasshopper for modelling, use the technology. Which in my view, is a fairly small, insignificant amount of people. Not to mention they do not require modelling tools, just code blocks. 

Inventor just recently received support (in 2016, for 2017 version) and the only time it's mentioned on their blog is in conjunction with an Inventor guy modelling the Fusion logo from one the Fusion tutorial. So knowledge and attention is limited, otherwise they would not have resorted to that tutorial...

Several posts on this forum seem to indicate that a small userbase uses t-splines, hence why improvement requests are not a priority to the devs. Alias was mentioned before, but I would like to add that Speedform received the most t-spline features, so I expect the team behind Alias to have been the ones to pull the cart. Possibly — and this is my speculation — the Alias division may have even been behind the acquisition of t-splines, if not then it was to avoid competitors running the toolset.  And now Alias seems to be pulling out...  So in addition, it would be nice to hear from the Alias team...

[cekuhnen]

It seems that they realized that T-splines simply has some foundational issues. I chatted over the course of t-splines growing up a lot with one of the product owners.

t-splines is also not the only technology out there that can turn Süd- mesh into nurbs.

considering that Alias is for car models I would simply say that they found t-splines not suitable being unable to deliver the surface quality and speed they needed.

regrading the user base - I find those comments always difficult. I do not model in t-splines because I find the integration just not as good and productive that sub d modeling in an external app and using t-splines just for converting.

i am quite sure if t-splines would work better as in a way many people users here suggested then more users would use it too.

would you ride a bicycle that lacks a pedal ?

[Intuos5]

---

@cekuhnen wrote:

regrading the user base - I find those comments always difficult. I do not model in t-splines because I find the integration just not as good and productive that sub d modeling in an external app and using t-splines just for converting.

i am quite sure if t-splines would work better as in a way many people users here suggested then more users would use it too.

would you ride a bicycle that lacks a pedal ?

---

Spot on, I would use t-splines more if some tools were to be added. I am currently undecided as to whether I continue to go along the same path of using a polymodeller and conversion method, for sure I'll be investing in learning one to apply the fundamentals. 

For sure the issues plaguing the design history with t-spline forms and a lack of capturing edit form operations it's workspace timeline (as a parallel to Edit Poly modifier workflows) do not positively contribute either. 

Also I don't think it helps that only 'create form' allows one to access the environment, which requires a different conception of modelling (and training) putting of a portion of the userbase. As has been voiced earlier in other forum posts, I agree that there is a fundamental lack in t-spline modelling tutorials to the point where I am first and foremost looking with suspicion as to whether someone makes 'mistakes' or not, before properly following the tutorial. The amount of disinformation is just shocking imo, so I am glad that some of you who participated in this topic spent effort in creating good tutorials!  E: It's a shame that the 'bad' tutorials are propagated by the youtube algorithm, thereby concealing some of those gems.

If — and it's a big if — Speedform were to be dropped, can we at the very least see some of its tools being ported to Fusion? 😬

[cekuhnen]

@Intuos5


Well the timeline parametric modeling in fusion cannot be changed.
Alias offers a design history which however is object based and not time based

Plus no app sofar captures the modeling steps in sub-d mesh.

[I_Forge_KC]

[TrippyLighting]

---

@Intuos5 wrote:

If — and it's a big if — Speedform were to be dropped, can we at the very least see some of its tools being ported to Fusion? 😬

---

How would you interpret this ?

One of the main limitations of T-Splines is that you can crease 100% or none.

The other limitation is that curvature continuity does not work at extraordinary vertices, star points in T-Spline lingo.

That however is a very important limitation that the team that has implementation Sub-D in Alias has overcome.

[cekuhnen]

Throw T-Splines out and replace with a better foundation like in Alias.

Deformations is something fusion in general needs not just only in mesh mode.

[Intuos5]

---

@cekuhnen wrote:
@Intuos5
Well the timeline parametric modeling in fusion cannot be changed.
Alias offers a design history which however is object based and not time based

Plus no app sofar captures the modeling steps in sub-d mesh.

I presume you are not familiar with the Edit Poly modifier/ modifier stack in 3ds Max/ Blender? I won't say it doesn't come with its own flaws, but at least edits can be made to geometry in groups of operations. Editing anything preceding the current Edit Poly modifier resulting to changes in the amount of polygons breaks the modifier stack aka 'timeline' in the way I refer to it now in Fusion. However, when used correctly, it works as like a timeline in which previously created edges can be moved and changes propagate to the later Edit Poly modifier. This makes it easy to delete a set of operations and to revert back to a previous state as well. See this tutorial to get an idea about how it works, excuse me for not having found a more to the point example.

I don't see a reason not to implement something like this in Fusion, but I am drifting off-topic here. 

@TrippyLighting I hadn't googled it yet, my bad. Two points come to mind in this. One is that the Fusion team knew this was coming and decided not to pursue a dead end. The other is that I fear it may take a long time before this sub-d method is implemented in Fusion. From 3ds Max, I know that there is a distinction in Catmull-Clark (support loops) and OpenSubdivs (creasing), so the OpenSubdiv algorithm may address the creasing issue you referred to. 

[cekuhnen]

@Intuos5 I might know Blender quite well being with it since 2001.

the stacks are coole but that is like feature in the timeline on a per object basis

Actually look up generative node based modeling in Houdini!

[TrippyLighting]

---

@Intuos5 wrote:

---

@cekuhnen wrote:
@Intuos5
Well the timeline parametric modeling in fusion cannot be changed.
Alias offers a design history which however is object based and not time based

Plus no app sofar captures the modeling steps in sub-d mesh.

I presume you are not familiar with the Edit Poly modifier/ modifier stack in 3ds Max/ Blender? ... I hadn't googled it yet, my bad. 

---

You probably should google @cekuhnen's YouTube channel to find that he is a master of the Modifier Stack in Blender and dare I say I am pretty skilled with it as well 😉

While the modifier stack adds some parametric abilities, it is not per-se a timeline.

I don't see reasons not to implement this, but then I don't have to pay for the resources to do it.

---

One is that the Fusion team knew this was coming and decided not to pursue a dead end.fear it may take a long time before this sub-d method is implemented in Fusion. 

---

We can of course speculate but having been to a number of meetings with the Fusion 360 team I don't think there is any cross pollination between these teams in a multinational multibillion dollar corporate organization. Alias and Fusion 360 are very different in ability and price point.

---

@Intuos5 wrote:

---

...so the OpenSubdiv algorithm may address the creasing issue you referred to. 

---

I've been asked that before but Blender works with Catmul Clark and you can still apply variable creasing all day long.

OpenSubdiv does offer this but I believe it is not yet implemented in Alias based on. forum thread. That might be incorrect on my part, however. More investigation needed.

[PhilProcarioJr]

@cekuhnen 

"Plus no app sofar captures the modeling steps in sub-d mesh."

Houdini does, just saying...

[Intuos5]

I guess I shouldn't have messed with the Expert Elites. 😛 

Apologies, must have been my phrasing that made it unclear. Rather than implementing a modifier stack, I can imagine an option where users add an Edit Form feature to the timeline which captures the history as opposed to all edits in the edit form workspace clogging up in a single feature (within the edit form workspace). That's how I could see similarities in functionality and nothing more. 

With that said, I'll follow along to see what others have to say regarding the opening post.

[cekuhnen]

@Intuos5  Na no worries - your blender comment was funny 😉 my students laughed.

Actually the mesh modeling feature in the timeline might seem good but keep in mind how quickly the timeline will be full. Plus that would be the mesh manipulation after which you would need to tell fusion to convert the TS to BREP.

But I fully get what you mean, like afterwards changing the subdivision of a TS revolve.

While Maya has a node based modeling I often found that with mesh editing this is problematic. Houdini is king at it but the complexity of the node at one point gets to confusion.

@Anonymous Yes I mentioned Houdini afterwards already.

[cekuhnen]

@Intuos5 

While I think the current implementation is a good balance between all options (super parametric models are also extremely stiff to manipulate when very complex)

I think what really would be killer is to throw this timeline and and parametric modeling stack in general under the bus.

To me a much more visionary approach would be to use a node tree to create data sketches surfaces mesh.

While complex too the bigger it gets however the advantage is that it would be more generative parametric modeling.

But I hardly think Fusion is going that direction and that is also fine.

[TrippyLighting]

Yep, and that node tree is exactly what makes Houdini so powerful!