View Full Version : principles to STEAL - ZBRUSH2

jin choung
04-28-2005, 04:10 AM
howdy all,

long time no see. been busy with life/work, etc. hope all you guys are doing well.

anyhoo, been playing around with zbrush2 and there are definitely some cool things about it that we should implement in future iterations of lw.

notes (#3 is the sexiest one imo):

1. zbrush2 doesn't seem to be super advanced in terms of software architecture... i can definitely see it being similar in complexity to lw, especially how it treats subdivision surfaces (basically as polys). BUT, it allows HIERARCHICAL editing (you can dynamically model at any stage of subdivision).

i always thought that having a history capable of dealing with such editing would be beyond lw's capabilities but it doesn't seem so anymore.

should we strive to incorporate it?

2. our METABALLS (etc) should be geared toward usability with SDS and not as a creator of a FINAL MESH... that is, the really really REALLY nice thing about zbrush is that all of its tools are geared toward creating SDS FRIENDLY meshes!

EVERYTHING comes out quads for further editing as a subdivision surface.

metaballs etc are basically USELESS for most purposes because they tesselate into triangles. but they would gain REAL USABILITY if they made quads and allowed resolution settings suitable for creating sds cages.

3. INTEROPERABILITY - zbrush allows edge weighting but also allows you simple slice closely spaced edges to create sharpness on sds so as to allow such meshes to work CROSS PLATFORM on other apps like maya and lw.

EDGE WEIGHTING on SDS (as i have ALWAYS argued ;) ) is ESSENTIAL for a really good sds implementation.

but we must realize ONE VITALLY CRUCIAL IMPORTANT POINT - edge weighting and other such exotic technologies are NON TRANSFERRABLE!!! edge weighting in lw (when we should get it) will not be transferable to MAYA's edge weighting or C4D etc.

UNLIKE NURBS, there is no STANDARDIZED implementation of SDS and so only the most LOWEST COMMON DENOMINATOR FEATURES can port over (by which i mean to say that you can export the raw cage to be subdivided in the target app).


we should do zbrush ONE BETTER and implement a feature by which EDGE WEIGHTED MESHES can be recalculated such that all weighting information is REMOVED and instead, appropriately placed close spaced edges are cut in to approximate the weighted version!

then, we can use the sexy edge weights when we stay in lw but we can also use a non weighted (but identical looking) mesh should we have to use a multi-app pipeline!



Chris S. (Fez)
04-28-2005, 05:11 AM
Hey Jin. For a while we were both pushing pretty hard for hierarchal subds in Lightwave...then Zbrush 2 came out with the implementation I more or less imagined, only better and faster than I thought possible.

"appropriately placed close spaced edges are cut in to approximate the weighted version"

I think it is a great idea. I would also welcome the ability to weight edges with negative values too, which I suppose could also be approximated with some sort of "relax" operation.

But to be honest, for me personally, the main appeal of weighting edges is that it does away with excess geometry.

Though this is all assuming Newtek adds edges...

04-28-2005, 05:24 AM
Add fast interacton with high polygon objects to that list.
In ZBrush, I can handle many many polys without slowing down the app to a crawl.
IŽve tried a model in ZBrush at level 6. Could handle it without problems.
Imported the lower resolution cage (level 4) into LW.
Opened the mesh, which took minutes, and viewport interaction (just polys, no subpatches) was practically non existant. At level 4, ZBrush was 6.
I have to append, no other application I tried comes near ZBrushs handling of high resolution meshes.

Honestly, I have no hope that weŽll see anything of that in LW anytime soon.

ThereŽre tons of minor things LW doesnŽt handle ATM :(

04-28-2005, 06:38 AM
Hi Jin! I think the idea #3 is gorgeous, and achiveable (well, since Im not a programmer, I must think so:)...). I have used Modo for some time, not Zbrush, but I can tell the edges things are so important for the evolution of Modeler and our workflow speedness. The #1 applys very well too

04-28-2005, 11:23 AM
I think modeler should be able to do what z-brush does to with paint on geometry add and subtract brushes.

Modeler never reall has had an in program 3d painter. Its long over due and yet would be cutting edge compared to other programs. Plus if it was native in modeler the integration would be perfect as opposed to needing plug ins adn what not.

04-29-2005, 12:01 AM
i don't know how this is realized in z brush, but guys Zbrush can handle millions of poligons , and has a memory management wich is very optimized. remember that the basic idea of Zbrush is that it's a 2,5 D app. so, i would guess that many of the things we see when we deal with huge models, aren't treated as real geometry in that moment : for example could not exist when nwe don't see them on canvasuntil the moment the geometry is saved, and maps are created. i maybe wrong on this tho. in this case, it means that zbrush is so well optimized in terms of memory allocation and geometry construction and positioning in 3d space that Lw and all other apps are way behind it.

however : a more z brush approach could be possible, but... it would need a lot of programming and good ideas.

04-29-2005, 12:12 AM
I agree. From what I understand zbrush isn't a true 3d program, and if you tried to do the smae thigns in a true 3d program it's be much slower. Zbrush doesn't even use hardware acceleration and look how many "polys" it can handle. It really is an amazing program. So while I'd like zbrush tools in lightwave (I hate the zbrush interface and I never know when I've flattened an object) we'd have to accept that it'd never be as fast.

jin choung
04-29-2005, 01:13 AM
howdy guys!

glad to see you fellows still oot and aboot!


actually, i DON'T believe that we should follow zbrush down the path of SCULPT MODELING....


BECAUSE ZBRUSH2 does it SO WELL already!

certainly, it would represent a VALUE for lw users if lw could just incorporate ALL the nice things about zbrush but this is something that i would argue is not really do-able in a realistic sense. same thing with 3d painting imo.

if you need to work at detail levels where you need to model extreme high frequency detail, i think the best solution is simply to use zbrush2 and then bring that detail back into lw via displacement/bump.


the 3 points that i mention tries to recognize that RADICALLY altering lw is not really an option and i think that they are features that are in line with what lw already is/ trying to be.


lw, maya, modo, xsi, etc. all represent technologies that are geared towards generating models comfortably to a CERTAIN LEVEL OF COMPLEXITY.

whether you're using polys, nurbs or sds, there was always a level beyond which we (and by we i mean everyone from prosumer to ILM to pixar) just resorted to a 2d paint app to create bump/displacement detail. (of course more recently, we had the advent of bodypaint, deep paint 3d, amazon paint, etc).

zbrush2 just takes that to a new level and allows you to sculpt in traditionally painted details on the geometry itself - that being desirable because it is more predictable than having to second guess how a 2d map is going to translate into 3d. (actually, it's really a product of the current ZEITGEIST along with the technology to bake geometry into normal maps for games like doom3, farcry, etc.... the epitome of the zeitgeist actually, i think).


having played around with zbrush2, i would say that it does function as a totally 3D APPLICATION.

yes, you can make 2d art.

yes you can make 2.5d art.

but you can also make fully 3d meshes that are exportable to any other 3d app via .obj.

and yes, even when a mesh is FULLY 3D, zbrush2 does a REMARKABLE job in pushing around even millions of polys... so we can't just brush that achievement under the rug by saying that it gains such advantages by not being fully 3d.

but it is true that when you want to sculpt EXTREMELY high frequency detail, it is much more responsive to paint 'above' the mesh in 2.5d and then 'stamp it' onto the mesh (which gives you all the advantages and results of 3d modeling while avoiding the pitfalls and slowdowns).

anyhoo zbrush2 is really cool and if you need to work at extreme levels of detail, i think that's the only way to go. and y'know what? that's ok!


p.s. but there are things that traditional modelers are much better at doing than zbrush2... zbrush tends to ABSTRACT much of the mesh away from your control... that makes it easier for the newbie but harder for the pro... so for instance, you CAN end up deleting a poly in your mesh and creating a hole but then NOT be able to make a poly to fill it and make a 'solid' again!

and even though the zspheres is a pretty cool way to sculpt something out roughly, it's still faster, easier and more predictable to generate your low poly cage in traditional modelers like lw etc and then take it into zbrush if you need the extreme detail.

i'd treat it as a NICHE solution... not really the best at general purpose modeling.