Model fillets...

Groboto creators coded Meshfusion...dont know how this will effect their own package though.

Well, in principle one could make this work comparable to the way Indesign references text documents and images or a Video editor references film clips. The best comparison probably was the assembly workspace in a Solid Modeller, if you are familiar with these. Essentially one could bring in Geo from any SubD app and live-link it. Inside Groboto one could move, rotate, scale, shear, replicate and do all sorts of other transforms on this geo. In case the mesh topology needs to get altered, one could do these operations in the source app and re-save. This could give anyone a pretty comparable feature set as available through the mesh-fusion plugin, even including their graphical node representations. The question is whether the Groboto makers have time to do this and whether the Foundry lets them continue the Groboto project.

The Foundry has some nice clips on their Website, this one is a comprehensive introduction too.

Thanks for the link. It is certainly a big improvement on Groboto. It as got my attention, so will have a look.

Simply amazing.  I have the impression this is something really big to program and could justifiably require a full version upgrade of Nvil with additional with upgrade costs.

I am playing around with Modo, and still feel more comfortable with Nvil simply because streamline tools make the basic modeling workflow much smoother and the context sensitive menus, tools, hotkeys really keep things clean.


A powerful boolean modeler is still an important workflow though.  Perhaps the community knows a GoZ to modo workflow.

I have the impression this is something really big to program and could justifiably require a full version upgrade of Nvil with additional with upgrade costs.

Currently in Nvil, it can take several minutes for a base boolean operation. Other vendors stopped attempting to code for booleans(can be very complex) and instead starting using the Carve CSG library. Blender now uses that library, and with the inclusion of Bmesh, now gives boolean results without all the triangulation.

With Meshfusion, that is technology from Groboto, but also uses technology from Modo, so quite a few years of programming by numerous programmers. Not something I can see one programmer doing.

Blender now uses that library, and with the inclusion of Bmesh, now gives boolean results without all the triangulation.

No triangles? Are you sure?

The Foundry has some nice clips on their Website, this one is a comprehensive introduction too.

From the video, I notice two things

1. The meshes used are all subdivided. I wonder what's the result for low poly models.
2. I didn't see the end result editable mesh it produces. Can it do that?

http://www.youtube.com/watch?v=T-ccEk3P5g0


Drool...would love these kind of advanced modelling tools.

I can see triangles in the result.

Basic workflow is this:


The plugin smooths the surfaces in order to create a catmull-clark surface so that there are no triangles.  Then, under the hood I suspect it is converted to a nurbs surface but displayes the tessilated surface in triangles.


Because it is nurbs, you can have the fillets and realtime performance.  Groboto is entirely in nurbs which makes sense for this plugin because they are just converting their tech.


Once the user is satisfied with the result,  the nurbs surface is then converted to a polygonal mesh.  The only visible triangles are along the seams of the boolean.  The results are satisfying though, and will definately change the workflow of a lot of modelers.  Booleans are essential in doing hardedged hiPoly work.   Very little clean up is needed, and the results are much better then using Zbrush.


As for the subdivision modeling for hard edged type stuff...Triangles and Ngons are not a problem...and can actually be exploited to help speed up workflow and keep geometry light and clean.

Blender now uses that library, and with the inclusion of Bmesh, now gives boolean results without all the triangulation.

No triangles? Are you sure?

Triangulation is only made on non-planer faces(but they have to have quite a large deviation). Otherwise any triangles are from the actual mesh boolean positions(triangles from the overlapping geometry).

From Carve CSG:-

Apart from requiring planar faces that are not self intersecting, Carve places no restrictions on the form of input data. Faces that make up surfaces may have an arbitrary number of edges, and are not required to be convex. Carve does not produce triangulated output, which protects against errors caused by poor triangulation that can impact subsequent CSG operations.

I have been using the Carve CSG boolean library for quite a while, as it was/is used in Wings 3D > ManifoldLab.
From my own point of view, triangulation made on booleans just add lots of work cleaning up/removing the triangulation. I prefer the approach (as with Carve CSG) where a cut is made on the intersections of the objects, and no extra edges are added to the geometry.

Maya is using it as well since the latest release.

While nobody knows about the details, I'm pretty perfectly certain that MF doesn't work Nurbs-based interally. Nurbs are far too complex for calculating all these realtime intersections, even by skillfully hand-modelling in Nurbs some Surface Blends are very hard to solve.
There are some automatic Mesh to Nurbs converters but all of them are FAR, FAR slower and most of them can't at all handle surface-trim, which is the most essential thing in Nurbs Boolean operations.
Also it made no sense to convert to Nurbs and not give access to this geo but only to output (less precise) meshes. I am pretty certain that the Booleans are performed based on triangulated, dense meshes internally, far less likely in voxels. Afterwards they are creating another mesh on top, which in its topology is more conform to what mesh-modellers would expect. The Groboto-makers have worked for years on this type of meshing,in particular of the blend areas. In the presentation clips one can see that the blend-quality of the merged mesh relies on the initial subdivision of the input-meshes, this also speaks against the use of Nurbs. Nurbs don't even appear in Groboto, what is used here are Quadrics. Fillets/blends in Groboto are a post process which gets introduced at export time, they habe nothing in common with Nurbs fillets.
Istonia: That whole process is not designed towards low polygon modelling. Also I have not yet seen any Boolean Engine which can tesselate the outcome without triangles and / or ngons, let alone one which allows for interactive updates. All meshers in the market which can output just quads and edgeloops (i.e. in Zbrush, 3DCoat and Mudbox)do this expensive and time consuming post-processes.

Also I have not yet seen any Boolean Engine which can tesselate the outcome without triangles and / or ngons, let alone one which allows for interactive updates.
All meshers which output just quads and edgeloops do this expensive and time consuming post-processes.

Blender 2.7RC, with its history, you can move the objects for interactive changes in the output boolean. On quick tests, for 2 object that are made up of approx 30,000 faces, I am seeing a feedback/update at about 3 FPS.

From Carve CSG:-

If triangulated output is required, we recommend that it be performed as a post processing step on the final result. Alternatively, hooks exist in Carve to allow a 3rd party triangulator to tesselate faces before the final result is produced

So it is exactly the opposite, as triangulation is a post process.

Polyxo,

Groboto is most definately using Nurbs or similar math.  It is all spline based geometry and then converted to polygons afterwards.  What is required is working with all quad geometry, which is why Mesh Fusion must smooth the geometry first. Cad modelers use similar methods and have no issues calculating fillets and booleans with realtime performance.  Even Maya can do this, although manual setup is tedious.

Groboto is most definately using Nurbs or similar math.

It uses "Quadric surfaces".

>Some info<

Groboto is most definately using Nurbs or similar math.

As I said: It's using quadrics primitives, which is analytic geometry but not Nurbs. They use some quick realtime mesher comparable to those used in CAD-applications,(in order to render a shaded display)  at export they use another mesher, which lets users introduce blends.

Cad modelers use similar methods and have no issues calculating fillets and booleans with realtime performance.  Even Maya can do this, although manual setup is tedious.

You are mixing things up. One may indeed create boolean operations with highly reliable results and relatively fast performance in CAD-applications. Multibody intersections may take some time though, just pushing a body through another one with the manipulator while having the geometry update was quite unlikely. The mesh which gets used to shade geometric outcome always contains triangles, only MoI can turn these to a mix of Quads and  Ngons at export time.

What one can not at all do in Nurbs, are these highly complex realtime blends as shown in the MF demo clips. Yes, there's fillets as parametric "features" in Solid-Modellers as Solidworks which can get edited later - but trust me - what is shown in MF can't be done in any CAD-Program. One can of course model all the shown items and with far greater precision. But one can not do this with such liberty and not with instant blends on most complex surface transitions.