Search - User list
Full Version: metaball modeling
Root » Technical Discussion » metaball modeling
deecue
so i've been messing around with the idea to model my ‘character’ in metaballs..

it came to me the other day when i thought of how people in the history of art used circles, ellipsoids, and spheres to draw and sculpt things. (i.e. traditional 2d animators makes their characters based off them, michelangelo as well as others use to use them to construct the muscles of a body, sculpturs use them as a unit of measure for a rough model)

so basically i am using metaballs with different shapes, sizes, and weights to get my rough model. then i use a convertmeta sop to convert it to poly's at a low level of detail. i then use the poly's to get more rough detail in. i subdivide a little bit. add detail with the new create subd's. repeat if necessary. and finally subdivide again for a finished smooth model. it seems so far to be a very fast way of working and i kind of like it. the only thing it really is good for though is more organic stuff. i.e. you couldn't model a bridge very well this way.

so yea, no one around here has ever heard of modeling this way so before i get into it too much, i figured i'd ask around and see if anyone has attempted this method before and would like to share their dislikes/likes and benefits/drawbacks to it.

so far i have only encountered something that maybe someone has a solution too (but i think it's just the way it is): so, i have my metaball network merged and converted, edited, subdivided, edited, and subdivided. if i go back to one of my original metaballs and just rotate it in the slightest, my point numbers go crazy. i actually completely expected this to happen but was wondering if anyone knew a work around.

hopefully this inspires someone else to try the method if they had not heard of it before either. but im hoping to find people that have as well. and also a work around to the point numbers would be sweeeet.

so yea, thanks for listening to the long post everyone

dave q
Dave_ah
Isosurface modeling is really passe method. I never liked it, and do not reccomend. Its more trouble then its worth. You have to set up these complex spline networks then Copy(SOP) metaballs into them, using splines as stamp surfaces.
You are better off using sub-d's to model your character then use spline metaball networks as magnets for muscle rigging. THis has been my R&D subject for past 8 months.

Dave
deecue
Hey Dave,

Thanks for the great inputs. I actually had never thought about stamping the metaballs to a complex spline network via copy sop. It's an interesting approach, but like you said, it sounds more trouble than it's worth.

I was more just using a multitude of metaballs laying next to and within (intersecting) eachother to mold it in to something. Almost as a sculptor would use various size spheres of clay layered on top of each other and smoothed down. After I have a pretty good rough model, i convert it to poly's and then to sub-d's. Then i can edit(sop) to refine the sub-d's and after that, sub-d again, edit those for detail, and do a final sub-d pass for smoothness at the end.

Here was an example i did last night. This is the rough metaball model of the head for a full grown (9 month) fetus i am working on. It was converted and sub-d'd twice. No edit sops for detail were implemented yet. It's not great by any means as it is the metaball rough, but it was an advantage to have that basic head model done from begining to end in ~3.5 hours. And if i need to make adjustments to the overall mold of it, i can do so very easily by going back to the metaballs. Comments welcome as always…




So now that they are technically sub-d's, i think i might take your idea with using a spline metaball network as magnets for the muscle rigging. I was looking for a way to do it without doing a bone rig since this isn't really a character thats gonna be walking and talking. just slightly expanding/contracting legs and arms.

Do you think that this method is still not such a good idea to go with? I definately appreciate any kind of advice on it.

Thanks,
Dave Q
JColdrick
I would agree with David…a far as whether is passe or not, well… I remember for a while there was a package that purported to be The Next Big Thing In Modelling based on that concept a few years ago. You never hear about it anymore.

The thing is, metaballs are great and all, but they're mathematical constructs(well, I suppose everything is, really), and you'll spend most of your time struggling with radii and metaball weights. I'm not sure I equate that with sculpting, but to each their own.

Learning to model polys with an eye toward subdividing is much more intuitive, much *much* faster in terms of computing power, and what's more you'll be learning a lucrative skill. Really not trying to shoot down a way you might like to work - certainly if you find it useful then use it, but the speed issue alone(computing 50 metaballs vs maybe 200 polys isn't even in the same ballpark) would make a huge difference.

Cheers,

J.C.
Dave_ah
Well if you gonna pursue isosurfaces then you should be intimately familiar with anatomical musculature layouts. Thats where you get out the ol' copy of Gray's Anatomy. Look at the curves that muscles follow and draw the stamp curves to conform. Then you need to write a script that alters the radius of stamped metaball so that the rad. of starting and ending ball is small and large in center, and intermediate balls have their radii properly set. As I said, its more trouble then its worth, and you spend to much time mucking around with raddii then thinking about shape.

Now for muscle rigging. You have to have a normal bone skeleton. The stamp splines have their end points tracking the attachment point on bones, as in real muscles and ligaments. So by normallly animating the skeleton the muscle curves deform properly (i.e. priocedurally). The metaballs are stamped onto those curves. Finally the MagnetSOPs read use the muscle isosurfaces to deform the skin.

Dave Rindner
tallkien
This technique was a passing fad a few years ago I remember when I was using MAX. Tools like Meta Reyes and Clay Studio Pro were introduced on the same lines claiming to build models from the inside out so to speak(muscles to skin). They soon went out of style for the very reason you mentioned. There was never any control over the final surface. If you left the skin dynamic, the point count would go all over the place, and UV mapping? …unless you project testures, dream on. Or the static solution which gives no control over the flow of the mesh. Try skinning a character generated this way and you'll understand. These solutions just didn't make sense for animatable characters. Not for me me anyway. (P.S. I do not mean to bash these tools, they are powerful indeed, just not the best way to go for me)
…And then you have the inflate Sop. Have you tried playing with it? I imagine it could be used as a modelling tool although I'm not sure if its worth all the trouble of setting up-just for modeling anyway. I might consider it as a study tool( to experiment with anatomical variations maybe- character OTL idea lurking somewhere in there?)

SubD's are the way to go for now though. Smooth finishes and complete control over the flow of lines on the surface. If the character is being animated it's important to have points exactly where you want them so as to deform properly. IMO Iso-surfaces are better suited to non character models
deecue
Hey guys,

Thanks for all input. It is definately appreciated.

Talkien: I noticed you mentioned the inflate sop.. Someone over at odforce mentioned that, but i cant seem to find it. Is it new to 6.1?

Thanks,
Dave
JColdrick
nflate sop.. Someone over at odforce mentioned that, but i cant seem to find it. Is it new to 6.1?


Yup.

Cheers,

J.C.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Powered by DjangoBB