< without turning it into polygons?
Yes :)

-> What would there be gained by applying sub-d smoothing

Hi Michael, most of the time I don't touch the topology exported by MoI but for example sometimes I find I need sub-polygon displacement on certain parts and I haven't planned at the start.

Marc

Hi Flashfire,

I see you have gone for really low poly output. That can cause issues as already mentioned by Michael, but you do also have bad flow, which causes problems if sub-d was to be added.

I did not think you would actually attempt that, so I have cut up the model in MOI to show how I would do this. This is at the minimal I would cut the model up, although I would probably do this directly in a Poly modeller.



If I was then to add sub-d(for whatever reason), the model would maintain shape, although I would add hard edges to keep boundaries.




- Steve

Hi Marc,

> but for example sometimes I find I need sub-polygon
> displacement on certain parts and I haven't planned
> at the start.

Could you maybe show an example so I could understand what kind of displacement you're talking about? Do you mean something like painting on some ridges and raised areas?

One method that can work well for that is to import your model into a brush based displacement program - 3D-coat has a voxel method for sculpting like this, which is independent from the initial mesh topology since it converts from a mesh form into voxel form for doing the actual sculpting.

- Michael

Hi Michael,

>>Could you maybe show an example so I could understand what kind of displacement you're talking about?<<

Some applications still use micro-poly displacement based on quads(a direct high sub-d). I know, some(including renderers) now use voxels.

- Steve

Hi Michael,

I was thinking of images maps or shaders like noise applied to a surface in a displacement channel, a bit like a bump or normal.

Here's a example on simple cylinders that has been tweaked a little to keep their original shape while subdividing.

Marc

Hi Marc, but doesn't that render-time displacement that you're showing work on any kind of polygon model, not only one made up of just of quads?

In fact in your example there, it looks like the cylinder to the right has an n-gon cap on the top, not all quads on the top, yet you show it displaced in the rendering?

As far as I can tell, you can do "sub-polygon displacement" in Cinema4D on any kind of polygons - that means to dice any polygons up into smaller pieces before perturbing their vertices by the displacement, the "sub" in this case doesn't mean "subdivision surface" in the same way as a HyperNURBS smoothing.

Here's a quick test just to verify - here's a model that is not suitable for sub-d smoothing, it has not been retopologized into quads only, it's got just 1 big n-gon polygon on the top:



Yet it can still be rendered with a displacement maps with "sub-polygon displacement" enabled:




However, since displacement mapping works in Cinema4D by perturbing vertices around (instead of by perturbing micro polygon render samples which is a way some other renderers operate) you probably want to export such models out from MoI into more finely diced up pieces.

You can do this just at meshing time though, you don't need to retopologize your model into a subdivision surface friendly topology first, you just use the "Divide larger than" option when generating a mesh to dice big polygons up into small pieces (and you may as well just use Quads & triangles instead of n-gons), that looks like this:



Having more regularly spaced vertices like that will make more points available for the displacement mapper to push around. But note that this is just an option that you set when exporting the mesh from MoI, you don't need to totally reorganize the topology of your model for this purpose.


There are some render techniques that do combine displacement maps with subdivision surfaces as well (not sure if this applies to Cinema4D in particular though), but it is typical to be able to do displacement mapping on just regular polygon models. Just generate a diced up mesh if your renderer doesn't incorporate displacment directly into its sampling pipeline like RenderMan.

- Michael

Hi Marc, so in other words you can use the "Divide larger than" export option in MoI to subdivide (as in "dice up into smaller pieces") the generated mesh if you need a more even distribution of vertices.

That kind of subdivision though is not the same thing as producing a "subdivision surface" friendly topology.

- Michael

Hi Michael,

Like said previously, this is a situation were I end up wanting displacement and is wasn't planned, or that it would very long to produce the geometry in a polygonal software.

Here are other examples with the cylinder,
When dicing very small, the vertex normal breaks make funky artefacts.
When dicing very small and unwelding vertices small openings appear.






Anyway this situation doesn't happens often and I'm a happy camper.
Just experimenting a bit. :-)

Thanks for the hints!!

Marc

Hi Marc, you'll probably want to use a welded mesh so that the pieces don't jiggle apart.

Also are you just using the default UV coordinates for that displacement mapping? That's probably what is causing your unwanted disruption at the seam between the top surface and the side.

You'll need to get a continuous UV mapping on your shape if you want the displacement map to also behave as a continuous looking thing across the whole model.

If you were to apply a regular image map to that piece, that would probably show you a similar break between the pieces as well.

I think it's likely that the UV mapping on your other piece is the reason why it behaves how you want, not particularly something to do with subdivision surfaces...

- Michael

Hi Michael, I didn't thought of unwrapping in the previous example, I used the default cylinder mapping because the default Uvmap stretched the texture.

I will try this later on.

Marc

Steve Wow! How did you force Moi to make such uniform polys? I mean I tried to slice but had no luck
like that.

Hi FlashFire,

Sorry, I missed your reply/question.

For that model I just used lines to trim the surfaces. For some models it can take longer to split up the surfaces than actually building the model. But if that is the type of output wanted, then that it what I do.

I did have a look at Topogun due to Michaels post (I had been meaning to look at that program for a while). The program is very easy to use, but find that it would take a lot longer to produce the type of output I want, than it takes me in MOI


- Steve

I tried to trim the model using the models outlines, but I don't get your results.
I've used both a re-join surfaces approach and non rejoin.
Could you show me what am I doing wrong?

You need to add lines to use for trimming.

Here is the start I made on that last model, I added a circular array of lines to trim the top surfaces. It is then a case of adding more trim lines where/if needed to make more cuts. It all depends on the surfaces if trimming is needed, as some surfaces will have good flow for the mesh output even without trimming.

But how do you decide on how to trim just any model?
I've been poly modeling since 1993 ;)
I followed your example......IT doesn't magically do what you did.....

>>But how do you decide on how to trim just any model? <<

Thats like asking how do you decide to build a model. It depends on the model and output wanted.


>>I've been poly modeling since 1993 ;)<<

I was never really into poly modelling back then. I did try 3d studio and used truespace for a while when it came out, but I was modelling in CAD at work, so that became an hobby when Rhino was first released in beta.


>>I followed your example......IT doesn't magically do what you did.....<<


If it magically did, then I would not need to add trims.
The circular array was just the start, I added more trim lines to cut up the model. It took me about 30 minutes to add the lines/cut up, so its not very quick, but still quicker than some other methods.

I noticed the more I cut the model using the radial method, the worse the polys jumped
around, as you can see in both images I posted. If you could please show me what you did
to get such perfect polys, I could then hopefully see your method ;)

I wish this area of Moi could be worked on to be even better. But realize this is a limitation
of the libraries used in programing.

Hi FlashFire,

> I wish this area of Moi could be worked on to be even better.

Could you be a little more specific and narrow it down a bit - which area in particular are you referring to and could you describe a bit about what you would rather see MoI do?

But please don't give a description like "just emulate my extensive experience and use my same judgment in creation of the mesh", because unfortunately software algorithms can't be easily created that follow such nebulous types of processes. A software algorithm has to follow a very specific and meticiulously detailed set of rules to produce output, it can't "just do it" ...


> But realize this is a limitation of the libraries used in
> programing.

It's not so much just a library limitation - it's that MoI cannot easily just repeat a process that is mostly based on human judgement and experience.

But there are other issues beyond that as well - one of them is that an all quad topology is not actually better 100% of the time.

Here is a pretty simple example:



So note there that the current mesher placed a single n-gon polygon on the top cap there. If you were looking for the cleanest possible wireframe and most efficient polygon count, that current result delivers that. To produce an all-quad topology for that same example would involve tiling that top planar face with a lot of little quad fragments, making a higher polygon count and denser wireframe instead of a really light wireframe.


MoI's mesher is oriented towards those kinds of goals currently and not on producing an all quad ready-to-subd mesh which it sounds like you are looking for.

If you want an all quad, ready-to-subd mesh, I really recommend that you create that from the start in a polygon modeling program that has a bunch of tools that are oriented towards creating that kind of a thing. MoI's tools are not oriented towards that kind of a thing, you can create poly output with MoI, but it is a different kind of polygon structure than that. The result from MoI is perfectly suitable for rendering directly already though.

If you're trying to do something else than render the polygon result from MoI you are probably just not using the correct tool for the job, except in certain special cases.

- Michael

Nah....I'm not wanting perfect quads.....Chose Moi for it's hard surface and boolean ability.
I wanted to point out that sometimes poly artifacts can show up even when C4D is used to render a moi model imported with normals.
Many points have to be manually welded/cleaned up.

Look at the polys to the left of the words output: and display:
If the exporter could weld points which are as close as these show, it would be even nicer.
Or some how handle these close points or sliver polys better.
This model is using CentroidTriangulation=n as y is so much worse. With Y you then have many points in the center of polys to weld or remove in a polygonal modeler.




In example 2 these points are so close they would no doubt create a smoothing problem in a polygonal modeler. So they would be welded by hand.




I should mention I have adjusted both divide larger than and avoid smaller than with many settings tests and still get slivers or
points that should have easily been welded to a neighboring point.

3D modelers I've used such as C4D still produces strange results at times using N-gon output, so I'll steer clear of it and settle
for Quads and tris. I hope this maybe helps.