Page 2 of 3

Posted: Thu Jul 08, 2004 5:22 pm
by Money_YaY!
Dani wrote:

BTW, is there a way to "implicitize" an object? I've seen this (i think ;) ) in other software, Truespace... where any object can become meta...

Dani
Ys I second this ! That would make the tools of edge loops and super fast and detailed modeling so nice.

You could create the head first separate from the body in high detail and and a body later and even mix and match the bodys. This would save so much time spending to stitching them .

Posted: Thu Jul 08, 2004 5:32 pm
by intrr
philovivero: Very interesting stuff. I've tried to model humans with metaballs before, and I've had similar degrees of success, but I lack some basic knowledge of human anatomy to make them as good as yours.

About the edgeloops thing... Don't spend too much time thinking about some comments... some people just like to throw buzzwords around :)

Posted: Thu Jul 08, 2004 7:02 pm
by theeth
If you don't convert the metaball to meshes, edgeloops don't really matter. If you are though, then you'll have a harder time getting a smooth armature deformation since metaball poligonization will not give you a mesh with geometry oriented in the sens of deformation.

Martin

Posted: Thu Jul 08, 2004 9:37 pm
by myamee
philovivero: nice work. I have a free version of Amorphium 1.0 and I just want to know if it utilizes metaball. Here is a female done in Amorphium 3.0
Image
This is more organic modeling than meta balls, but I just thouht the two are related some how.
MyAMeE

Posted: Fri Jul 09, 2004 1:42 am
by philovivero
If you don't convert the metaball to meshes, edgeloops don't really matter. If you are though, then you'll have a harder time getting a smooth armature deformation since metaball poligonization will not give you a mesh with geometry oriented in the sens of deformation.
Yes, of course. I am talking baby steps here. First, I need to be able to model a static human. I don't want to use meshes. It's like assembly language programming when Java exists. Sure Java is slower, but you develop faster.

To model a static human, I do need some way of telling a metaball which other metaballs to blend with. If this is done with directing structures (this would be like skeleton, right?) that would do. The earlier objection "How to keep forearm from blending with upper arm when bent?" doesn't matter, because forearm would blend with elbow with upper arm.
I have almost finished some important part of "bounding box" optimalisation.
Oh, this is great! It will speed things up and help dramatically. I'm still concerned about my realisation that this doesn't stop the forearm from blending with the upper arm, which if I can get "editable directing structure" might solve the problem.

I am wondering if anyone else has had success modelling say a HEAD in one .blend, a BODY in another, then putting them together into one .blend? I would be willing to do this, but haven't figured out how to move complex structures between .blends (or to make layers work).

Posted: Fri Jul 09, 2004 1:54 am
by philovivero
Jiri, I think you wanted to do mesh-guided metaball systems. Maybe using armatures to give a hierarchy, a meta only blends to his parent and to its child.
This seems brilliant and simple. Is it possible?

I went to look at directing structures on your page, Jiri, and I don't see how they solve the problem of "don't let hand blend in with shoulder" whereas Dani's armature hierarchy sounds good.

Except, well, somehow I have to be able to attach one metaball to the armature bone, and another metaball to the armature joint. Can you do that in Blender presently? (Attach certain vertexes to bones vs. joints)

Posted: Fri Jul 09, 2004 2:20 am
by Money_YaY!
Download the demo from Zbrush.com there is a windows and mac version. Learn how to draw Zspheres, I am sure this is what you are going for. Not getting someone to create this in blender is a different matter.

Posted: Fri Jul 09, 2004 10:23 am
by bjornmose
Yup inttr interesting .... idea:
Such a metaball based object could be ideal as "inner" collision constraint for soft body clothes. I mean not reandering the metas themselves but only the clothes "covering" them. .... skin could be done the same way .... and if cloth cares for self collisions the arm forearm issue is solved. Do you think this works?

BM

Posted: Fri Jul 09, 2004 10:49 am
by oin
Some comments:

I started many years ago modelling. And when I got to the point of organic meshes, thought first of metaballs. As is the more inmediate way for a traditional clay modeller person.

But is not the ideal way, imho. Anyway, it can be useful, but you will wish also to know other ways of modelling.

I lik ethis thread for see shots very familiar to me.


I started , collaborating a bit with the author of Softy. I helped him a bit with suggestions. And indeed I had a pair of models in the gallery a human head ( only found this render, someone asked me to use it in a render he was making) and a human arm...Not very accurate. Indeed, too much polygons to aquire precision, if u ask me.(metaballs in general)

Softy is metaballs and soft splines, like metashapes guided by an spline skeleton.

I had not all the precission I wanted. So I moved to Organica. Was almost perfect workflow: a quick metashapes builder like it, allows many more shapes and quite control on each shape and strenght. It even had a basic skeleton system. Could even do some anim. The exported into Amorphium 1, which I bought too, and rocked for mesh refining.Even more, it's feature, Memshman, allows to paint in grayscale, and the near to black areas are optimized, near to white keep the count, with all the subtle tones inbetween.Great to have with a metashapes modeller.

But still, long proccess, and hard to get accurate ness. Tried splines, nurbs with Rhino demo. Then made models with Hamapatch. To constrained. Then the solution came : subdivision. Polygon modelling and subdivide. That was the thing. Found Wings3d and the problems ended :)

Today, I have AMorphium 1 and Zbrush 1.55b purchased (as well as deep paint3d and Ultimate unwrap) just in case I really wish a very detailed ending. But all the modelling I make it with Mirai style modelling of edgeloops and subdiv that Wings3d gives you.

If I want to make a normal map, or just a hi res model vwith veins, cracks, etc..Amorphium and Zbrush do rock for this finishing stuff.

It also can be done directly with metaballs modelling..but you'll end up with really high res meshes and still not too accurate... I mean, the time to do it is imho bigger than build the basic low pol cage model and go making ctrl tab in Wings3d to preview the subdiv version. (or work in proxy mode)

Of course, finishing touches with Amorphium 1 (somehow I prefer it to Zbrush for this, weird but is a matter of likes-dislikes...Besides I don't need a new serial number from the company each time I reinstall my OS(often)...) or Zbrush...


BTW, it's been my path (not mentioning a load of other tools inbetween)
but it could be very different for you.Minds work differently and luck could be other.
Any modelling method can do if the idea is in the mind ;)

But if u try other methdods at same time, may save u years, maybe.

And sorry my poor english...

From mesh to metaballs

Posted: Fri Jul 09, 2004 11:27 pm
by myamee
Money_YaY! I had downloaded this avi yesterday from http://e-learning.vslib.cz/~hnidek/anim ... taball.avi . My stupid question is what was the button was selected to turn the mesh in to metaballs.This was the only thing that I did not understand.
MyAMeE

Mesh as directing structure for MetaBalls

Posted: Sat Jul 10, 2004 12:36 am
by jiri
Hi all,
myamee: this video is captured from my personal devel version, so I guess you don't have this button in you blender ;-)

philovivero: bone of armature can "deform" every object including MetaBalls, but I affraid about scatterd MetaBalls :-( ... you will see

You cannot simply say: this part of MetaBall will not blend with another part of MetaBall, because surface of MetaBall is evaluated using IMPLICIT FUNCTION!!!
Polygonisation algorithm doesn't solve if some point of surfcase "belongs" to the first MetaElem or second or some other MetaElem. Point of implicit surface belongs to every MetaElem in scene. Is it clear now?

It is possible to make some model in one blend file (using metaballs) and then append this model to another blend file ... File->Append (select .blend file) ... Object ... select objects using RMB ... click on "Load library" button :-)

"implicitize" an object? ... intrr's soft body will do this, i hope ;-)

btw: Ton commited some edge data structure to the Mesh today, which will help me implementing my idea: mesh as directing structure for MetaBalls, thaks ton :-)

I hope, that I answered all your questions ;-)

Sorry for my poor english.

Jiri

Posted: Sat Jul 10, 2004 3:39 am
by TorQ

MetaElems are not Holy Grail.

Posted: Sat Jul 10, 2004 11:07 am
by philovivero
Oin: Thanks for your first-person account of going down the path of Metaball organic modelling. That's a very useful account.

Jiri:
You cannot simply say: this part of MetaBall will not blend with another part of MetaBall, because surface of MetaBall is evaluated using IMPLICIT FUNCTION!!!
Polygonisation algorithm doesn't solve if some point of surface "belongs" to the first MetaElem or second or some other MetaElem. Point of implicit surface belongs to every MetaElem in scene. Is it clear now?
I might understand. Let me say it again and you tell me if I have it right.

If MetaElems A,B are in one scene, then A shall blend with B. If I add a MetaElem C into the scene to blend with B, there is no (simple) way to make C not blend with A, for they are all part of one scene. (One large shape)

All: From TorQ's post:
Yes! Look at the second animation! (brazo_la.avi) and see how the muscles change volume and shape depending on the placement of the limb. How do they solve this problem? It looks like they're using MetaElems of some algorithm or another. Do they just cheat and not show that if the hand approaches the shoulder, they shall blend together?

Different algorithm of polygonization

Posted: Sat Jul 10, 2004 12:02 pm
by jiri
Hi,
they propably use little bit different type of polygonization. I guess, that they use some explicit techniques ... generalized cylinder or something like this. We don't use this technique, because this has some limitation and can produce some discontinuites, self-intersections, problems with branching, etc. I doubt, that it could be used for some liguid simulation. Of cource using this technique you can say A will not blend with F, G nad X, but will blend with B, D and Z.

This technuque is much more similar to extruding along some curve.

Let have look at this paper:
http://www.unchainedgeometry.com/jbloom ... rfaces.pdf

Regards

Jiri

Re: Mesh as directing structure for MetaBalls

Posted: Sat Jul 10, 2004 4:05 pm
by Money_YaY!
jiri wrote:"implicitize" an object? ... intrr's soft body will do this, i hope ;-)
Jiri
It just makes the mesh squshiy. It does nothing to make it 'meld' with the other mesh.

Really if we could have this feature building models in peices would be so useful. It could mage freaks of nature or you could focus on part of the mesh at the time which would be great for makeing great details for normal map extractions.

Please look into this, you would be prasied for ever.