In working with some of my software (older Cinema 4D for poly, or Rhino for NURBS) it seems it would be nice to work on a system of multiple scales. Let me explain.

Lets take C4D for starters. I can start with a cube and make it editable. I can place the cube as a child of a sub-division (MetaNURBs, yeah I know, it has little to do with true NURBs) object. So now the cube wil be sub divided on render. I can select the 8 points of the original cube and move them to make largh scale changes in the shape. But I can not do the same with the sub division unless I apply or collapse the sub division object. Then I can make more five modifications. But, I have lost the ability to make the large scale modifications. So here is what I am asking. Are there any modelers that allow just such a multilevel control at the same time. THe ability to manipulate a few control points for large scale changes, but be able to manipulate sub divided control points for medium or fine shape control,.. all at the same time? Without collapsing anything? Seems that this could be a powerful, and intuitive, way to model. Especially if it could be done in at least 3 levels (large, medium, and fine scale control.) OF course, changes would need to be able to go down and up the nesting. It may be difficult, but I can not imagine that it is impossible.  Are there any tools out there that already do this.

The same idea of multiple scale control points can be applied to NURBs as well.

Henry

ZBrush allows you to move between a low and high poly subdivided version of a model (and levels inbetween) for a sculptural approach. I'm not sure of the exact terminology for this in ZBrush.  In Daz's Hexagon this is called "Multi-resolution smoothing".  I expect that Silo and Mudbox employ a similar method of working at different levels.

For more info

Hexagon
Mudbox
Silo
ZBrush

I don't own ony of these but I have tried ZBrush and was excited by the possibilities.

 It’s a nice idea Henry, and there’s certainly been more than one occasion where I’ve said to myself, “darn, I need to change the base shape of this but I’d really like to keep what I’ve done to it so far,” while I’m working on the detail, but I’m not sure how feasible it would be to implement a system that allowed you to do so for a few reasons. I think the way that apps calculate subpatches, and NURBS, would prevent this from really working, I think it would cause point order/identity issues as well as create a serious slow down in performance.

Most apps that have subpatching/NURBS capabilities use an algorithm something along the lines of taking two point and the points surrounding those two, and using the angles/distances of those points to derive a “tension” of the curve between the initial two points, and that tension gives you the curve you get when subpatched. What that leads to, for one polygon refined once, is 4 pseudo-polys and 9 total points, 5 of which are “projected” or estimated based on the initial four points.

To implement a system that allows you to modify the 5 new points, while still having the first 4 control the overall shape and such, you’d need some method of tracking those 5 points that would uniquely identify them so that they keep whatever changes you’ve make to them. What I think might be a problem though, since these new points don’t truly exist they’re just approximated based on the initial 4 and thus relative to them.

I think that it might be difficult to track all these “child” points seeing as none of them are “real” and all relative to the “parent” points, and they’re all sort of “stacked” on top of the first 4 points. Should you decide that the jump from “large” to “medium” is instead of 1 polygon into 4 [mesh refinement of 2], is one polygon into 9 [mesh refinement of 3], now where’s that leave your modified points from the first pseudo 5? Because, now instead of 4 real and 5 pseudo-points defining 4 polygons, you’ve 4 real points and 12 pseudo-points defining 9 polygons, and any modifications you made to the pseudo 5 is now likely lost, or needs corrected. When you move to the “small” level, again you’re creating even more pseudo-points at an exponential rate, all of which need to be tracked, none of which exist, and they all are relative to their “parents” and any modifications you may have made to them, which are then in turn relative to the first 4.

What would work is using cages. It doesn’t necessarily give you the large-medium-small [L-M-S] effect you’re after, but a “deform cage” lets you manipulate the mesh as a whole based on a simplified cage and adjusts the mesh and details accordingly. Example being, you’ve got a 2500 point sphere and you want to make it wobble like Jell-O, instead of moving all 2500 points around you can use a basic cage that consist of maybe 25 control points. I think it’s more of an animation tool though and I’m not sure it can be used for modeling. I know XSI has a deformation cage, and I’d imagine Maya/Max have something similar as well as most other apps, LightWave excluded, at least prior to the latest release.

For some reason, I keep thinking Blender might have something along the lines of what you’re describing but I’ve not gotten into it deep enough to say for sure, you’d need one of the Blender users around here to answer that.

It may also be possible to work like you’re wanting using splines and patching them. Seeing as you could keep adding splines to increase detail in certain areas where you’d want to tweak the “medium” level. This would probably mean a lot of patching and un patching as you go to check how things look and what not as well as a real solid understanding of and high comfort level with splines.

If you had a way to prevent the point order issue, like a way of identifying points relatively say, polygon 124, is defined by points 112, 114, 118, and 120, then said the pseudo were point “124, A, 1” “124, A, 2” “124, A, 3” and so on you might be able to track the points while keeping them “soft” or not “real.” That would leave with having to just work out a way of eliminating the problems incurred by changing from a refinement level of 2 to a level of 3 or whatever.

Also, if you follow the “proper” workflow it should greatly reduce the need for the L-M-S, that workflow being “general to specific,” meaning you rough in the basic shape and gradually increase and define details as you go working on the mesh as a whole. I usually make sure to keep a lesser detailed/earlier version of any model around so that should I desire to make changes to what would be the “large,” I can and then only need to remodel a part of the mesh. Even as it is now, a lot of times you can make the changes you’d want with a different tool, like a “scale/stretch,” “twist,” “bend,” “shear,” or so on though it would probably be a lot easier in an L-M-S system. Who knows, maybe a properly implemented L-M-S system could go a way to redefine the approach to modeling as it is now, sort of like Zbrush and that approach has now.

DISCLAIMER!!! I don’t code, I know some BASIC, but I’ve never done much with it, so I could well be wholly wrong about what kind of slow down may or may not be incurred by this sort of system. I do know that the more geometry you have, the slower things get, and that having geometry that relies on other geometry, doesn’t help things any. Secondly, I don’t know all that much about point order or how different apps sort it, I just know it can sometimes be very important that you don’t mess with it. Lastly, by “proper workflow” I don’t mean to say that any method is incorrect, I’m just going with what I’ve been taught, and feel is the best method as well as what is generally accepted as “correct” not just in modeling but for art in general and of course not say that the ”proper” workflow doesn’t have issues either. In the end any workflow that gets you where you’re going is just fine. I’m rendering right now, and thus bored, and having a lot of time to run my mouth, so you can take any and or all of it with a grain of salt and like I said I could well be wrong about all of this, but this is my two cents on the matter so enjoy.

Swishy [I see we cross posted I rambled so much], is that allowing you to actually work and the details in at the higher resolution then decrease the res to a lower a level and still keep those high res details? Or is it just like adjusting your mesh refinement, from say 3 subdivisions to 6 or something?

If its the former, I could be eating my little essay for dinner.

Tho even at that sculpting modelling is a good bit different than "regular" modelling I believe. I've not had the chance to use Zbrush or the like, but I do know it relys less on mesh flow and more at just making mass/volume/shape, and then cleaning it up later.

Intriguing topic anyway. I hope we get some more ppl to weigh in on this.

Yes, you can work at the higher level and return to the lower level while retaining detail - at least in ZBrush and, I would imagine, the others I mentioned

The higher level is, in most cases, used to produce a normal or dispalcement map which can be applied to the lower level model. For a model in C4D that would be applied to the original base model dropped into a HyperNURBS with the required subdivision level. That's about the limit of what I've picked up.

I think you'd still need to pay attention to the basic edgeloops to avoid any unexpected deformations when animating/ deforming and it would be wise to model the major forms rather than trying to do too much with the sculpting.

It's very appealing but RGB normal maps only seem to be supported in the higher end apps at the moment.  They allow the surface of the model to be displaced in 3 dimensions rather than just a +/-% option of a greyscale displacement map.  They'd allow things like caves and overhangs in a terrain formed from a plane. In that respect they allow you to push the modelling to greater limits but how advisable that is I don't know. I'm going to have to wait until the technology filters down to the lower end before I'll be able to find out. :)

Any time I've seen a normal map applied to a "low" poly mesh, you can still see the edges and faceting of the base mesh, grant it, its usually pretty well disguised because of the way the normal map breaks up the values on the surface, but it doesn't alter the silhouette at all to my knowledge. Normal Maps as I know them are mostly just like juiced bump maps and as such and in conjunction with bump maps, can lead to some incredible levels of detail with relatively low polygon counts.

Displacement maps however, in LightWave's incarnation anyway, will actually alter the geometry, however, how good that looks in the render depends on how much geometry is present to be altered, and as such can lead to the need for astonomical amounts of polygons.

Swishy, you are correct about still needing to take into account the flow on the base mesh. I've seen workflows both where people will model the base mesh, sculpt that, generate the normals, and apply them back to the base mesh as is if its sound enough to begin with, and also where people will model the base mesh, sculpt that, generate the normals, then create a new base mesh over top of the high res model to make sure they have the right poly flow and silhlouette then apply the normal map[s] to the new mesh. 

As I see it, you wouldn't use the normals to create the cave, I think that would need to be modeled with actual geometry, but instead use the normal maps to make some awesome looking cracks and gnarly textures on the cave's walls.

It shouldn't be too long before the technology filters its way down to you. Normal maps are getting to be pretty mainstream now it seems, as well as sculpting brushes. In one of the more recent releases of Blender they've included sculpting capabilities which I played with for a bit and it looked like it promising, I'm not sure what the level of support is for normal maps in Blender at the moment tho, didn't get that far. As it was for me to get a working setup to create and use normal maps in LightWave I had to seek out a couple free third party plugins all of which were a year or two or better old, so I don't think the tech is that far off from the "low" end.

Here's a quick normal map test I did about two or three weeks ago when I was getting for me to create/work with normal maps up and running. I did it mostly to test the limits of how far you can push normal maps, and also because I'd never used them, just seen them used, and crashed LW anytime I did try to use them. 

The box on the left is the "high" poly [not that I'd consider it high, more of a moderate at 9500 polys],  I used it to generate the normal map [object space] for the middle box, which is 26 polygons. Then cut the high poly box up and used it create a bump map for the far right box, which again is only 26 polygons.

You can see both the normal map and the bump map break down, and you can tell that its an image affecting a flat plane when you get off to the side a bit. I pushed this test well beyond the limits of what you would use normals maps for and used them to simulate a lot more volume than they should.

The middle row is just the boxes with the maps applied to the color channel rather than the appropriate channel and the bottom row is both maps, greatly compressed as they're both 2048². I've not taken the time to compare an object space normal map to a tangent space normal map, but I don't think the results will vary that much. Perhaps that'll be somethign to look at tomorrow. 

Geometry will typically, but not always, look the best but at the price of having that many more polys in the scene and the associated render hit. Comes down to finding a balance of what needs modeled, what can be normal mapped, and what can be bump mapped, at least that's how I see it.

A very effective demo and I agree 100% with your last comment about finding the right balance.

Mudbox allows you to sculpt on layers so that you don't have to worry about screwing up the underlying mesh. Silo 2 also has an option to allow sculpting on the subdivision only or allowing the sculpt to alter the base geometry. Both applications employ a form of non-destructive modeling, so that you can choose if alterations affect the underlying levels before the one you're working on or not.