As some of you know, many poly based programs that offer sub division surfaces use the Catmul Clark algorithm and are mostly quad based. Some of these also offer micro displacement.

My thoughts is why couldn't that control shape that they use to create characters and organic models be regenerated or interpreted by AM splines?

In other words if it was possible to import sub-d models made by programs such as Hexagon, Blender, Z-Brush, Silo etc and create a control surface so that people can rig and animate them in AM.

Wouldn't this open up a pretty large market and give people that model externally an opportunity to use AM for rendering and animating? It seems that there are so many sub-d fans out there and they are shortchanged when it comes to rendering and animation.

well, briefly... that control cage creates topologies that are similar to A:M topologies, but not identical. For example, Sub-D models contain "poles", vertices that are corner to more than 4 polygons. Those instances create ambiguous spline continuity that must be resplined manually in A:M. That resplining always involves creating new CPs where no vertice was in the original model. This requires some judgement.

If you can come up with an algorithm that will always solve that correctly, and all the other "not identical" stuff, you'd be on your way. Remember that Sub-D cages are generally NOT the actual surface of the model. A:M patches ARE the surface of the model. Another difference that must be bridged.

But I don't think mere mesh import would be enough. Someone who has been working in z-brush has been making maps too. Those would need to be successfully mapped onto the new A:M topology.

Easy for our human eye to see what needs to be done. Very difficult for computer to see this and do it.

http://www.ghost3d.com/gt_ResErect.htm

Yes I am since day one jealous of that. I would like to have that algorithm for other software. If it would be sold for the other big players, I would in an instance buy it and use A:M for the rest.

See you
*Fuchur*

Then from 3DSMAX it is exported in IGES for any other software....
...
Possibly, this question would dare at presence plugins import FBX and COLLADA

Interesting. I'd never heard of this software. They state in their FAQ that it ignores 5-point patches. What about hooks, I wonder?

>They state in their FAQ that it ignores 5-point patches.

<then it`s quite useless.

The ResErect tool looks like a very clever thing, but as I look at the screen shots of what it does it is still making topologies that would have to be fixed in A:M. So using it to clean up a polygon model prior to getting it into A:M seems unlikely.

(BTW, I suspect it hasn't been updated to accommodate the V13 file format.)

Really, Sub-D models are already the least trouble some of any polygon models to import into A:M. They already have topology closest to what the A:M topology would be. But there's still that last 5% that has to be done manually.

The res tool looks like it could work.

My thoughts were in the case of sub-d specifically that the Catmul Clark has a constant smooth angle, never any creases or edges and I would think would be the easiest to convert. The base poly surface of the sub-d controls 2x2, 4x4, 8x8 etc number of quads within. AM's 5 point patches and hooks shouldn't be necessary since they relate to the multi influential cp. Models that have been exported out to say .obj, could be decimated and interpreted back to fewer patches.

Catmul Clark smooths don't have variable smooth angles or variations in sub division on a per model basis. I would also think that the model to be converted should be limited to a single continuous mesh at a time.

The only real issue I have run into with sub d conversion here is with shapes such as cubes or the end point of a quad because the obj plug in attempted to read the end patch as 2 loops from the sides.

I just see so many people creating with these surfaces and they have very few options when it comes to animation and rendering time unless they are ready to shell out some serious bucks.

There's the flaw in the plan. All complex shapes in A:M use and need those. Character shapes converted to A:M WILL need to use those.

Just as an exercise, try manually converting a Sub-D model into A:M without 5-pointers or hooks. Actually do this and show your before and after here.

I have done a Sub-D to A:M conversion before and animated the result, it can be done, but you need the 5-pointers and hooks to fix the unacceptable topology that Sub-D creates.

Yup your correct. I guess then the best solution would be to have a specific exporter made for each native application that exports out the control shape so that could be used for rigging. All the underlying geometry would have to be polygons generated from the original program along with the textures.

So basically there needs to be some sort of hybrid of spline controller and poly model.

Of these things, "it's possible... like going to Mars is possible," Martin says.

Actually just simple polygon support so that the modeler, even if you can't edit the model, be able to rig a poly model and support texture maps that have been pre-assigned to it.

This would certainly open up the doors for many new customers out there.

For simple, import the obj and peak all the splines. That would be simple polygon support. You could rig that.

What you want is not as simple as that solution, I bet.

The lack of substantial polygon model support in A:M is not because they didn't think anybody wanted it or never thought to try it or didn't know what polygons were.

Martin has put good minds on it before and they found it impractical.

But the SDK is out there, if it's really simple get a programmer on it and prove them wrong.

The problem with the obj importer is it still converts and attempts to create patches out of faces and that doesn't always work.

The prop importer seems to be the best option because it reads the polygons just as they are.
The problem with the Prop import is that it doesn't support multi material single mesh objects so assigning shaders to them is limited to 1 shader per prop and the imported textures are washed out (most likely the way textures diffuse, and specular value are interpreted).

Always frustrates me when I see an opportunity that is over looked.
I wish I had the ability to program this.

The washed out effect can be fixed once you put the object in a chor and go to the "Surface" parameters and set the "ambiance" to zero. This model had 3 surfaces, the skin and two eyes.

Click to view attachment

The head on the left has the ambiances corrected, the head on the right is the default.

I'm not sure where I got this head. It might be from "Make Human". And I don't know where I got that.

Click to view attachment

the "lo" version of that, imported as a model has a not-hideous mesh that could be edited into proper form with moderate effort, and rigged. Reattaching the bitmaps back would need some R&D.

the "hi" version looks better initially as an MDL than the "lo" but would be trickier to rig.

Thanks for the info on the ambiance setting.

When I import this obj file into AM as a prop, the ambience is at 0 but no textures show.

This model has 3 shading domains with 3 textures on a single mesh (single mesh multi material). This would be pretty much the same as a decal stamp on selected patches.
Single mesh-single material works ok. This might be a problem with some applications exporting out mesh groups. The single material single mesh standard for 3ds files.

As for rigging high density meshes it might be nice to have a brush selection tool that the user can drag over the surface. A similar tool option for selecting patches would also be very handy.

I know making these additions to AM would be time consuming but better compatibility especially on importing would increase the number of users. There are so few good programs that have the animation tools this has and so many poly modelers out there that could really take advantage of this.

The head imports blank to you? When i import it as a prop I get textures.


edit: you mean the sphere.

The head comes in fine because the head is separate from the eyes. If it was one single mesh when it was created and textured then it would come in blank.

If you have an obj viewer you can see the sphere test is just one mesh with 3 separate textures on it.
The point is that if people make a character in a sub d surface modeler they often assign multiple textures to a single mesh much the same way we decal in AM on groups. If they were to cut/paste those material groups into separate meshes the sub d surface would be junk and they would lose the continuous surface that the catmul uses.

So I guess it would work out like this... the prop would need to be able to support the multi mat single mesh from obj files and then we would need to be able to rig a prop even if it has a gobs of polygons.

Anyways its a thought.

The OBJ head imported as a model and rigged with bone weights. No attempt was made to fix any spline continuity.


Click to view attachment

Impressive! What's the patch count on that monster?

ya but thats not a poly object.... missed the point
Not all obj files will come in clean and most often need major rework to convert to splines. Rigging a poly prop would be the simplest way for super complex stuff.

That's a seriously dense mesh! Rigging the face on that beast would take some creativity ... Nice work on the neck though!
One interesting thing, even the parts where 5 splines converge on a single CP seem to look OK.

That was 6054 patches neé polys.


How may dollars of development time do you think it would take to achieve flawless polygon import including super complex stuff? Or how many hours?

Remember it has to be developed from scratch. It can't infringe on any other program's patents, it can't use any open-source licensed code that forces A:M to be open-source too.

Make a real estimate.

I think you could get something passable with bone weights. Maybe the clever solution would be to rig the lo res version and use Steffen's weight transfer plugin to move it to the hi-res version.




I think it's because each face is so small.

I must have been sleeping on the last couple of updates.....

The obj importer under this newest version of AM is pretty dang good. There are cases where triangulated edges pose a problem but if those are converted back to quads they translate pretty well. It also seems to read the multi material groups off of the single mesh (something the prop importer lacks).

Here is an old impact driver model I made a long time ago. The mdl file i fiddled with on the battery area and the obj. The model was originally from Hexagon and done with sub-d and simple poly shapes.

When I open the sphere I don't get textures. You do now?

Figured this out.... if you have an obj file you need to have named groups as well as textures. Single groups are not supported.
Not a big deal to do in most cases but I don't know how some of the programs handle this such as Silo and Zbrush. I know Hexagon can do this since thats where the impact driver was made.

Anyone here have Z-brush and willing to export out a test obj?

new sphere test zipped

z-brush offers a 30-day working demo... or did the last time i checked... you might get that and experiment.