Three Dee Editors

3D Studio Max (several thousand dollars)
blender
SmalltalkThreeDee (ST3D) which comes with a minimal 3D editor
WingsThreeDee
Maya
Alias
Wavefront
3dom
Amapi 3D
Arnold/Messiah
Art*lantis
Ayam
Aztec
Behemot
Bryce
Carrara
Electric Image
form-Z
goose Egg
Infini-D
Innovation3D
Inspire 3D
jPatch
K-3D
Merlin 3D
Milkshape
Ray Dream
Real 3D
Rhino
Shade
Softimage XSI
Softimage3D
Strata Studio
true Space
True Vision
Houdini
SmoothTeddy?
ZBrush

Almost no mention is made of ThreeDeeEditors for CAD (ComputerAidedDesign?) purposes. This strikes me as a grave oversight.

SolidWorks?
Pro/E
Catia
Alias Studio

The problem with 3D editors, apart from the incompatible file formats, is that they require 3D graphic artists with many years experience to run them. If you're an average person with modest needs, they might not be easily usable. Their makers may have never heard of InteractionDesign.

ST3D:

what is 'set UDD from normal' supposed to mean?
I still haven't been able to use more than half of the functions in the geometry editor
parts go completely invisible as a matter of routine
sculpting what you want is next to impossible

Blender:

some new users claim difficult to get used to
five rendering options available on-screen seems to count more than a way to select the object you want

Wings3D:

excellent
simple, reliable controls
obvious controls
the complete opposite of Blender; the makers of Blender acknowledge the difference
still Microsoft Paint and not Photoshop

What a 3D editor should do is to allow the user to sculpt a shape much like if they were using clay. The purpose of a 3D editor isn't to edit polygons anymore than the purpose of a text editor is to edit characters. The purpose of a 3D editor is to allow the user to define a SHAPE and to then COMPUTE a 3D mesh that reasonably approximates that shape.

For instance, the very first thing that any 3D editor asks the user before adding any kind of shape is 'how many vertexes do you want it with?' To which the answer is invariably "I don't know!" Do 3D editors allow you to redefine how many vertices will go in a shape? No they do not. Do they allow you to defer the OPTIMIZATION DETAIL of how many vertices will go in a shape? No, they do not.

What a 3D editor SHOULD allow you to do is the following:

drop a circle
drop a square inside the circle
join the two
declare it to be a surface
declare the circle to be an outer boundary and the square to be a hole
compute the polygons that go with that shape
compute more polygons in a designated area to make it smoother

The above example is rather unclear. Is the desired effect similar to what you get with boolean in Illustrator (i.e. a 2D shape in a 3D space), or is the desired effect a sort of donut with a squarish center? Also, the UserStory given above does not sound at all like working with clay.

Also, speaking of working with clay, there are at least three modellers worth looking at for different user interface styles. The first is Zbrush (http://pixologic.com/home/home.shtml). It is not just like working with clay. It does allow you to just "paint" out extrusion and "paint" more detail onto the geometry.

The second modeller that I want to point out is SensAble?'s ClayTools? (a piece of their FreeForm? system). This is a very expensive setup that involves custom forcefeed-back hardware that actually attempts to be true to clay modelling.

Third, Takeo Igarashi has created a free modeller called SmoothTeddy? (http://www-ui.is.s.u-tokyo.ac.jp/~takeo/java/smoothteddy/index.html) that is suitable for children to use. It makes models from the user painting sketches.

You're talking about a very, very dated way of doing things. All current modeling is generally done with some form of patches, or NURBS, or both. Maybe it's not working with clay, but it's certainly above polygon counting. Also my experience is with LightWave and a little Maya, I have no idea what the lower end/open source modelers are doing. For all I know they *are* 10 years behind in functionality. -- CallumLerwick

No, you, sir, are outdated. Subdivision surfaces are now extremely popular, and bezier (or assorted spline types) patches are hardly ever mentioned anymore. NURBS fall somewhere in between and are now often reserved for MechanicalComputerAidedDesign? or non-organic objects. Subdivision surfaces are highly similar to working with polygonal models since they are largely manipulated like polygonal models are. However, the tools for manipulating them have improved in recent years. The improvements tend to be associated with Subdivision surfaces, but most of the new tools could be applied back to plain polygonal models with no trouble. Some subdivision surface modellers are really polygonal modellers that offered a subdivision smooth function without allowing you to manipulate the low resolution cage while seeing the high resolution surface. Thankfully, this seems to be being fixed fairly quickly.

Of course, if their marketing copy is to be believed, LightWave 7.5 does subdivision surfaces, and is also just like modeling clay!

LightWave 3D introduced the first real time subdivision surface modeler and now the industry has embraced subdivision surfaces as the way of the future. With LightWave you own a robust tool set for manipulating these revolutionary surfaces, including interactive modifiers that allow you to model in multiple "soft" modes. LightWave allows you to sculpt "digital clay" by direct manipulation at the vertex level by controlling groups of points or polygons with an extreme list of operators.

From http://www.3dcafestore.com/lightwave3d.html -- CallumLerwick

Yes, I do believe that lightwave was the first popular commercial tool to have subdivision surfaces. I find it extremely amusing that they used to (and maybe still do) call it MetaNURBS, and originally were trying to use it to cover up the lack of NURBS. I don't know if Lightwave has ever added proper NURBS, but I don't think it really matters all that much. BTW, this feature of lightwave was introduced in version 5 I believe, which would have been 1995 I think. I'm pretty sure that 3DS Max didn't get subdivision surfaces until version 3, which I think came out 5 years later. I couldn't possibly say when Maya or XSI got it. For a long time I believed that Lightwave 3D was the superior modeller because of this feature.

And that brings up an additional usability problem for most programs, including Blender, Maya, 3DSMax, and Softimage XSI, which is that there are many ways to model. In blender, for instance, you can model using polygons, subdivision surfaces, NURBS, and lattice deformations (which seem to be effectively a hybrid between NURBS and subdivision surfaces). How is a new user supposed to pick in advance which type to use? Wings3D, ZBrush, and Rhino (and others) give only one method, but at least Wings3D and Rhino have major shortcomings to that method.

Well, that's the breaks when you're dealing with art. Some people like colored pencils, some people like watercolors, some people like airbrushing, some people like PhotoShop and a tablet. Some do them all. Can you argue that one is universally better than the others? Not really, it's entirely dependent on what your goals, subject matter, personal tastes, and desired style and esthetics are.

Advanced artists don't want their options limited in the name of "simplicity". It seems certain that trying to accommodate "advanced" and "casual" artists in the same artistic tool can only result in doing neither very well. -- CallumLerwick

Good point. However, I think that is an argument for having multiple programs for modelling, not trying to stick every modelling paradigm into one program, even if that program is meant for only advanced users.

I don't think that users particularly want to deal with those sorts of choices, and I do think that users want to be able to work in a more obvious streamlined work flow, and that Zbrush and SmoothTeddy? (which is probably not at all ready for professional use) are likely the first steps in that direction.

BTW, for many tasks, especially organic tasks, NURBS suck. NURBS to not allow for localized complexity in a model. This means that if you want to add detail to the face or hands of a model, you need to increase the complexity of the entire model to get the complexity needed to add the detail. The way around this is to actually split NURBS models up into many objects. For instance a model of a person might have one model for the torso, one model for each leg, one model for each arm, one model for each foot and hand, one model for the main of the head, but separate models for the ears and nose, etc. However, joining NURBS models together is not easy to do by hand, so more advanced NURBS modellers give tools to make it easier to make joins seemless, but that still causes complications, and now every vendors tools work differently.

Yeah so I hear from my friend who's learning Maya. However his school was still using Maya 5 last I talked to him...

'''I don't keep close track of what changes with every version

However, for Mechanical CAD tasks, NURBS may be superior as I get the impression that they work better in the context of CSG modelling.

"Artistic" modelers/renderers are something totally different from CAD/CAM stuff. And I know pretty much nothing about CAD/CAM. Other than CAD/CAM uses solid modeling, as you're modeling a solid object after all. Artistic 3D tools have remained pretty much surface based. They're shallow and superficial, caring only about appearance, if you will... -- CallumLerwick

Unfortunately, artistic 3D users increasingly have to worry about volume and the physics of things as well, because convincing animations require that the surface account for the movement of the mass underneath the surface, while convincing renders require accurately accounting for surfaces like glass, marble, and skin, which effectively requires accounting for how the volume reacts to light rather than just the surface.

If I were to go out on a limb and guess at the major difference between CAD and art users, I would guess that art users just need a model of a volume while CAD users also need to know how to create that object using real work practices. I would imagine that it is easier to figure out how to algorithmically create something represented as combination of NURBS and CSG than it is to figure out how to make something represented by subdivision surfaces. For instance an extrusion followed by filleting would probably be easier to convert to instructions for a CNC machine than a raw surface would be. But, as I said, I'm really going out on a limb here. It may just be a matter of historical forces.

Another issue that separates CAD and art users is the matter of measurement. CAD users generally need precise real world dimensions and measurement. Art users generally don't care so much. JohnDougan?