Hi Will,

> (Unless of course Michael can create that ever-elusive all quads exporter!)

Well, one additional tricky thing is that even if I had that it would not necessarily be a universal solution.

When MoI generates a mesh, it creates the mesh vertices on your existing surfaces.

Even if the polygons were generated with an all-quad topology, if you then turned around and used those same quads as a sub-d hull and then did further subdivisions, your subdivided results are not going to follow the original NURBS surfaces directly, there would be a kind of additional "melting" or shrinking-down kind of modification happening.

That may be fine if you want a very melted kind of a thing, but I would not be surprised to see things like some bumps and undulations being formed that were not in the original NURBS model.

So it's just not at all certain that converting to an "all quad" polygon mesh would actually be completely equivalent to a seamless conversion to sub-d...

Of course it would certainly be useful in many situations though, I'm just not very confident it would be a totally magic conversion thing like it kind of sounds like you would expect.

- Michael

Hi neo, just a bit more information...

> the question is which one of the two make better use
> of exported geometry from MoI?

Check out a recent discussion on the comparision here:
http://www.3d-coat.com/forum/index.php?showtopic=2724

One thing that is pretty interesting about 3DCoat is its "voxel" mode, which works rather differently than a displacement map type method.

If I understand things correctly, when working in that voxel mode, it does not really rely on subdividing the base polygon geometry to add further details - instead basically the imported data will just "turn on" an initial set of voxels and then the imported geometry is not used anymore and you work with manipulating voxels only after that.

So when working in that new voxel mode in 3DCoat, there is probably not any advantage to having an initial sub-d model versus a MoI-generated model. In other words, that new mode of 3DCoat may be more generally friendly to MoI exported data.

When you are done manipulating things in the voxel mode, then it sounds like you convert it back into polygons by using the retopologizing tools.

That mode in 3DCoat is a pretty different way of working than ZBrush, and it may be an easier way to incorporate any kind of MoI model as the base object.

- Michael

Also a tutorial for going from MoI to 3D Coat to Cinema4D is available here:
http://www.3d-coat.com/forum/index.php?showtopic=2166

That's just for painting and not sculpting but it may be useful.

- Michael

Actually I'm surprised to see that all tris went into ZBrush that good! (I never saw that thread before!)

Certainly something to give a try!

Ultimately everything is a tri anyway it's just maintaining the wonderfully clean mesh from MoI upon import to other apps!

Maybe one day the 3D world will be all tsplines and we'll have the best of everything :-P

-Will

Hi Will,

> Actually I'm surprised to see that all tris went into ZBrush that good!

It seems that ZBrush is able to handle high density data itself quite well, just make sure to turn off n-gons at export from MoI, and if you subdivide a MoI model inside of ZBrush, turn off the "smooth" (sub-d) subdivision option.

Also make sure you export a dense initial mesh out of MoI as well, using the "Divide larger than" parameter to dice up any big polygons.


I think the main issue is you need to worry more about what you are going to be doing with this model after you are done in ZBrush... If you are trying to make a bunch of characters for an animation for example, it is going to kill your animation program if you have 10 million polygons for each character.

If you know that you want to have a light sub-d cage + displacement map result out of ZBrush, then you would not want to use a MoI model to go into ZBrush since you won't get that kind of a result out the other end.

But if you know that you want to have a dense geometry final result, then it can work fine. You'll kind of want to watch how much subdivision you do so it doesn't get too huge with this method though. But it is feasible to add embossing detail to a basic object while keeping things more like 100,000 to 300,000 polygons rather than in the millions.

- Michael

PaQ thanks for your reply, I'm aware of all that I was just hoping some kind of improvement have been done between Nurbs modelers and painting apps.
I just fund it counterproductive that I have to redraw a mesh with correct topology in top of the original is the worst part of the process in my opinion. So I wanted to know if anybody here use some different
approach.
Anyway within few lines you describes the hole process in a very informative way.

Could nPower's quad mesher be useful in situations like that?
http://www.npowersoftware.com/booleans/pboverview.htm

> (Unless of course Michael can create that ever-elusive all quads exporter!)

1st. workflow is to Save from Rhino/MoI as .stl. Then I open it with meshlab. I "unify duplicate vertices" and then save
it as an .obj. Then it can be sculpted Zbrush.
http://meshlab.sourceforge.net/

2nd. workflow http://alice.loria.fr/WIKI/index.php/Graphite/PGP

Both free but not a magic Bullet.

3rd. workflow COMING SOON

Will this on involves 3d-coat voxelsculpting>quadrangulated>converted to tsplines>converted to nurbs>SolidWorks>Modo. VOILA



>Maybe one day the 3D world will be all tsplines and we'll have the best of everything :-P

I have been dreaming about that day

Hi neo,

> Could nPower's quad mesher be useful in situations like that?

In theory...

But I actually have the code for that quad mesher included in the Solids++ geometry library that I use, and I tried hooking it up and found it to not be robust enough for general use.

If you look at the image in the link you posted a bit more closely you can see that particular example would be good for making a kind of pseudo-fillet along the hard edges, that part is fine.

But on the internal parts you can see that there are some irregular areas, some quads that are nearly shaped as triangles. This is not really a problem in planar areas like the case shown there, but if the surface is curved those kinds of topological irregularities would probably produce some bumps or undulations when they were used as a source sub-d cage.

It's pretty easy to get things like lumps in sub-d surfaces, here is a tutorial that may help to explain:
http://www.guerrillacg.org/home/3d-polygon-modeling/subdivision-topology-artifacts

It tends to take some experience and human intelligence to make good sub-d friendly topology choices, I'm not very confident that a completely automated tiling mechanism is going to be able to do the same job in all cases...

But certainly it would still be useful in certain circumstances - if your main goal would be to round off sharp edges that part would probably work well with that style of quadifying, since the topology is good exactly along the original edges - it is where things collide internally that make problems.

But basically, I would say to not expect a magic silver bullet comprehensive "works everywhere" solution from something like that.

- Michael

Hi neo,

> 1st. workflow <...>
> 2nd. workflow <...>
> 3rd. workflow COMING SOON

What is it that you're trying to accomplish with all these things?

If you want to create sub-d output as a final result, most likely your best bet will be to create your model as a sub-d model, using sub-d tools from the beginning rather than using a different toolset and trying to somehow convert it later on.

Otherwise, one route that looks pretty promising for conversion is the retopo tools in 3DCoat. But if you don't have enough time to invest in doing that then you probably won't really have enough time to do what you want overall really.

It is just not very likely that you will find an automatic push button solutions, to make a high quality model you do generally need to invest some time and gain some skills to make such things come out well.

- Michael

Perhaps if some spoke from real experience with the latest version of 3D Coat, some of these various comments would be seen as superflous.

A flow between MoI, 3D Coat, and, as for me, Carrara 7 Pro, seems very logical.

Certainly, for ease of use/cost/ within my needs, far better than the Z alternative!

Brian

>What is it that you're trying to accomplish with all these things?

Are You Being Sarcastic? If not here we go.

1st. workflow (Graphite) & 2nd. workflow (meshlab) was more like a tip to whom wants to convert tris to quads.
I do build physical prototypes which then I scan after that I refine the mesh with Graphite or meshlab for further use in ZBrush, Modo etc....(depends the nature of the project).

3rd workflow. Well better show these images by yolk than writing an essay. you can see most of the workflow I was describing earlier.


quadrangulate voxel sculpting.


Converted to Nurbs with tsplines


imported into solidworks, shelled (wall thickness) and groove added.
.
.
.
.
Now for "render friendly" output, Modo 401 will have a solidWorks importer. (but I'm skeptical about the mesh quality. I imagine the mesh that is saved with the SW file for screen display will be what 401 Imports. But this is still a nasty STL like mesh.)...time will show.

"If you know that you want to have a light sub-d cage + displacement map result out of ZBrush, then you would not want to use a MoI model to go into ZBrush since you won't get that kind of a result out the other end.

But if you know that you want to have a dense geometry final result, then it can work fine. You'll kind of want to watch how much subdivision you do so it doesn't get too huge with this method though. But it is feasible to add embossing detail to a basic object while keeping things more like 100,000 to 300,000 polygons rather than in the millions."


Hi Michael, I guess this is rhetorical (until I try it for myself), but couldn't I use the hi-poly export from MoI into ZBrush generate the normal/displacement maps and then using a seperate, clean MoI ngon export use that as the low poly base object and just load up the paint & displace maps?

Or would I need the low-poly model in there as well as a part of the displacement process? (I believe that's (unfortunately) the case since I think it needs the level 0 base to refer to for the map generation...)

I also don't now if you can use normal/displace maps from a all tris model onto an ngon topology either.


Just asking, being that I've never attempted to generate normal/displacement maps from my models yet...



Using hi / lo MoI exports would be fine - my probs have always been just getting the models into ZBrush in the first place!

-Will

PS - Thinking a bit further - the normals/displacements would come from the initial tris input vs the displacements and additions I add to the model within ZBrush.

The question then is can I project those maps back onto an ngon version of the model??

@Neo

I did a lot a test in that area too, but I give up :)
The solidworks wall thickness looks amazing Oo

@WillBellJr

Normals map and diplace are really different.

You can 'bake' an high res model over a lowres one. N-gones should not be a problem, but I never tried that.

You can use a software like x-normal (http://www.xnormal.net/1.aspx), where you can import a low res object, an hires one, and transfer the details as a normal map.
However, I'm not sure if x-normal will load the MoI .obj with the right vertex normals info. If not, then the result will be really bad.

More less every 3d package can do this normal transfer now, the only problem is to import the high-res model. (x-normal is really optimized to import huge highres models).

For the displacement it's an other story, because your base mesh has to be ready to be subdivided with a clarmull-clark kind algo. So you really have to prepare this base mesh carefully (-> sds/poly modeler).

You can for example use 3dcoat for this job too, importing your huge moi mesh, rebuild a new base one with the retopology tools, and then bake the details has displacement map.

@Pilou

The decimate tool looks cool ... but ... I need to see how the wireframes look before doing a party.
All the existing decimate solution create ugly mesh, topology speaking, the result if visualy perfect, but it's not the kind of mesh you can use for a character setup.

But maybe they have develop a magic one :)

>> Hi PaQ, just one thing to keep in mind is that not everyone uses ZBrush for exporting to a rendering program - some people want to generate STL data to produce a physical model.

Hi Michael, indeed, I didn't think about that when looking at the sofa example.

Waiting your result about Decimation master :)

Hi neo,

> Are You Being Sarcastic? If not here we go.

No, I really didn't understand what your goal was for all of this.

I mean, in some aspects it is kind of like trying to put a square peg into a round hole...

If your goal is to get the final output into SolidWorks, then that makes more sense to want to do these kinds of steps - I didn't know that was your final goal until your new message here though!

So does that final goal of export to SolidWorks also apply to what you wanted to do with that couch example you posted previously?

With that one you still may be better off starting with a Sub-d modeler and making the general folds and ripples in it by manipulating the sub-d point cage. Then you bring that into t-splines and convert it to NURBS there. Right now that seems to be the main workflow for that kind of a thing.

But that quadrangulate result you show after voxel sculpting is looking pretty good, that looks like it may be promising!

The nice thing about that voxel method is that it should not be sensitive to what you use as the starting shape - whether it is initial sub-d data or initial CAD data exported from MoI it should not make any particular difference because the final result is extracted from the voxel structure.

- Michael

Hi Will,

> Hi Michael, I guess this is rhetorical (until I try it for myself),
> but couldn't I use the hi-poly export from MoI into ZBrush
> generate the normal/displacement maps and then using a
> seperate, clean MoI ngon export use that as the low poly
> base object and just load up the paint & displace maps?

I don't really know enough about the specifics to give you a good answer on this one.

It all depends on how dependent those mechanisms happen to be on working on sub-d levels rather than dealing with geometry directly.

From PaQ's reply it sounds like you would be ok to generate a normal map with this kind of method, but maybe the workflow of how displacement maps are set up to be used are more integrated into relying on the sub-d refinement process and so would not work for this.

- Michael

>No, I really didn't understand what your goal was for all of this.
I mean, in some aspects it is kind of like trying to put a square peg into a round hole...

yup that's me, a designer not following the crowd...)


>If your goal is to get the final output into SolidWorks, then that makes more sense to want to do these kinds of steps - I didn't know that was your final goal until your new message here though!
So does that final goal of export to SolidWorks also apply to what you wanted to do with that couch example you posted previously?

It has a dual purpose really.
a) create manufacturable data (in some cases as shown on example above...)
b) use as much of that data possible to create marketing material (photo real renderings etc.)


>With that one you still may be better off starting with a Sub-d modeler and making the general folds and ripples in it by manipulating the sub-d point cage. Then you bring that into t-splines and convert it to NURBS there. Right now that seems to be the main workflow for that kind of a thing.

Correct and that is what I have been doing all along. BUT I fund it counterproductive. (remodel just for the sake to get a "render friendly" output)
THERE FOR STILL IN SEARCH FOR THAT NEXT HYBRID.


>But that quadrangulate result you show after voxel sculpting is looking pretty good, that looks like it may be promising!
The nice thing about that voxel method is that it should not be sensitive to what you use as the starting shape - whether it is initial sub-d data or initial CAD data exported from MoI it should not make any particular difference because the final result is extracted from the voxel structure..

Indeed looks like it may be promising! Also I must say Michael Just your mesher alone has made a big difference...
All I see is Michael Gibson, Matthew Sederberg(tsplins) and Andrew Shpagin(3d-coat) each one holding a piece of the Rosetta Stone...) just put them together and we may have the answer to all questions.

I have yet translated two of these trio maybe I will ask the third :D
Thx for the idea :)