Modelling Heads

#41

j3st3r wrote:

Standard. Look at the final fantasy or osiris characters represented by Fransisco Cortina and Steven Giesler. Or look at René Morels AmazonSoul. They are relatively high resolution cages, I think theyre 100 000 polys. Quite much for a cage. But! The character I did for Platoon game cinematic is about 75000 polys control cage. The face had some 4000 polys, the rest gone on the cloth folds, tiny details, etc.

well all is indeed relative, but looking at the cages of for instance Fransisco Cortina,

F.Cortina Cage

it seems to me that the face/feet was initially modeled at a polygon level similar to that of the rest of the body, and then subdivided in order to yeild and extreem cage in those areas, I doubt that these areas (head, feet) where modeled using this many polygons from scratch (although that assumption is based upon my meagre experience.)

Splinegod wrote:

creating “end poles” thru triangle shaped quads.

ahh, that’s what it’s called!

The end use will drive the detail. At Weta for example they created very high rez versions of gollum which could probably never be used to animate. They generated normal maps that they applied to lower rez versions which preserves the details but not the outrageous amount of geometry.

yes, the use of displacement maps with lo-res cages seems to be an incredibly efficient way to add detail, atleast when looking at the tests that surfaced sofar concerning the upcoming ZBrush version.

although displacement mapping is nothing new, from what I gather it’s Zbrush’s ability to detail an enourmously subdivided mesh that is the key to being able to get the best out of the displacement maps.

unfortunately ZBrush at $399 is a bit steep for me currently so I’m hoping there will be cheaper (free?) alternatives somewhere along the line (or maybe Santa will help out ;).

and your Gollum is an excellent example of clean localised detailing in a mesh, superb modeling!!

#42

Originally posted by BinarySoup
[B]j3st3r wrote:

yes, the use of displacement maps with lo-res cages seems to be an incredibly efficient way to add detail, atleast when looking at the tests that surfaced sofar concerning the upcoming ZBrush version.

although displacement mapping is nothing new, from what I gather it’s Zbrush’s ability to detail an enourmously subdivided mesh that is the key to being able to get the best out of the displacement maps.

unfortunately ZBrush at $399 is a bit steep for me currently so I’m hoping there will be cheaper (free?) alternatives somewhere along the line (or maybe Santa will help out ;).

and your Gollum is an excellent example of clean localised detailing in a mesh, superb modeling!! [/B]

Thanks for the comments :slight_smile:
Actually its not really displacement mapping. Its a type of map called a Normal map. Its similar to the Microwave plugin for LW. Theres also a free Normal mapping plugin on Marvin Landis site. Basically the normals data for a high rez mesh is converted to a map that is applied to a lower rez mesh thus creating the illusion of greater detail then what is really there. :slight_smile:

#43

As I look at the models of FF or FFoO, I`m sure that at a certain level of detail they were subdivided, and then they applied many of final local details. Why it is good? Because when you achieve a generally equal polygon distribution over your model, it will be maintain during subdivision. Then addig local details would make no harm of the distribution.

I did it when I made my troll…

My site and the last page

#44

OMG!

AMAZING SITE!

Those models are beautiful!

I’m definately getting really excited to see how you guys put everything together.

#45

Very nice asian head, j3st3r!

Tris are inevitable … but I did note the triangles (especially the one that is attached to another) on the forehead of the second head. It looks like they could be easy to rearrange. Have you thought about terminating the splines into 4 point “triangles” in this area? That’ll get rid of the tri’s and give you all quads :slight_smile:

Killer website, by the way!:buttrock:

#46

.i use the point by point methood. i can build a head in a couple of days plus some time tweaking that way. *HARDCORE. :slight_smile:

but anyhow, heres a nice “head” thread from the previous newtek forum.
probably not much on technique or approach, but you can still appreciate and disect all the nice headshots. :slight_smile:
http://forums.newtek.com/discus/messages/2/17979.html?1028046416

good memories.

#47

I take some risk to jump into an LW discussion :slight_smile: and say a few things about Francisco Cortana’s FF models…

First of all the Animatrix models were using the existing pipeline, obviously for budget reasons.

Now, about Square’s pipeline, the important thing to know is that they’ve originally planned to do the whole FF movie using Maya and Maya only. Renderman only came to the party after completing a few shots (so the first 3-5 shots of FF are actually rendered with Maya - wow!) and at that time many things were set in stone - for example, they had to build Renderman shaders to exactly duplicate Maya renders, and they had lots of poly geometry, whereas PRMan is more efficient using parametric objects like NURBS.

The characters were under production at that time too, and the important consequence here is that the character pipeline was built with Maya’s restrictions in mind. A very important problem with Maya versions 1.0 to 4.5 was smoothing or subdividing polygons, because the app has not smoothed the UVs. I think that something similar existed in LW as well… the problem is that when you have smaller and bigger polygons near each other, then the UVs of the small polygons will get some ugly stretching. This is obviously unacceptable in full-screen views of the characters.

The solution for Square was to completely avoid sudden changes in polygon size (or edge distance) when building the characters, and this basically ruled out the use of any modeled wrinkles. It also resulted in a relatively heavy control cage, because of the need for relatively small polygons at the lips, noses and eyes. This has also lead them to use animated displacements to create the wrinkles for facial animation, which did not look that good IMHO.

So, this is the background for the looks of the FF character wireframes as far as I was able to put it together. The irony is that Gollum’s control cage is also made from polygons of very similar size - but it is almost twice as dense as the FF characters and thus wrinkles can be modelled into the blend shapes :slight_smile:

#48

A WITCH! A WITCH! A MAX USER HERE!!! ALERT!!!

Nice to see you here, Laa-Yosh!

You are wrong. This UV problem didn`t exist in LW. It exists still…
And how the things look like, that it will exist in v8 as well. Dense control cage can solve partially this distortion.

Chikega. The only reason I didn`t care with those tris, that they are hidden under a rigid mask…:))))

#49

TRIANGLES! MISCHIEF IN DISGUISE!

#50

Originally posted by j3st3r

You are wrong. This UV problem didn`t exist in LW. It exists still…
And how the things look like, that it will exist in v8 as well.

It’s good to know that LW is not the only program to have had that problem. But why be so certain that it won’t be fixed in version 8? You can always hope :shrug:

#51

Actually LW is the only one program that have this distortion problem. All the rest handles it perfectly (including Max, Maya, XSI, etc)

As I think it`s because the rest of the progs works on history-based environment. In LW it would be necessary to recompute the UV vertex position during subdivision. Like smoothing the UV.

#52

Oh yeah, I realise other programs don’t have this problem now. I was just pointing out that at one time it was a problem (according to the above post by Laa-Yosh ), that has been successfuly resolved. It’s about time LW did the same.

#53

great info there on the ff/animatrix movies there Laa-Yosh, suddenly those extremely dense polygons in the face area made sense, as for the Gollum character, wasn’t it 3d-scanned from a sculpture and then cleaned up? thought I read that somewhere, but then again, one reads a whole lot… either way, an extremely dense mesh for a character like Gollum that is to look realistic in numerous close-ups next to live actors in a film is likely warranted. btw I saw your cages over at metagons, superb stuff, I was totally impressed with the way you did the legs, now thats polygon efficiency and perfect loop disposition all rolled into one!

#54

Yeah, thanks agian guys, this thread has turned out GRRREAT!

Lots of info.

Keep it coming. I’ll start my head soon, the inspiration from this thread is amazing!

#55

Originally posted by BinarySoup
great info there on the ff/animatrix movies there Laa-Yosh, suddenly those extremely dense polygons in the face area made sense, as for the Gollum character, wasn’t it 3d-scanned from a sculpture and then cleaned up? thought I read that somewhere, but then again, one reads a whole lot… either way, an extremely dense mesh for a character like Gollum that is to look realistic in numerous close-ups next to live actors in a film is likely warranted. btw I saw your cages over at metagons, superb stuff, I was totally impressed with the way you did the legs, now thats polygon efficiency and perfect loop disposition all rolled into one!

Its pretty much done that way at most larger studios. Maquettes are created, scanned and then cleaned up. Gollums various facial poses were created by scanning LOTS of sculptures.

#56

In the LOTR DVD they show a box modelling sequence showing how gollum has been modelled. It is a quick animated GIF.

#57

Originally posted by telamon
In the LOTR DVD they show a box modelling sequence showing how gollum has been modelled. It is a quick animated GIF.

What blew me away was that the original gollum in cinefex was scrapped after a couple of years of development. The original was more amphibious but was changed to look more like Andy Serkis after Peter Jackson saw his performance.

#58

It is quite well explained in the DVD appendices. It is amazing, they have sculpted dozens of gollums. And once Andy arrived the casting session and performed, they understood that the traits of gollum had to be in close relation with his voice.

In the appendices you can also see the Keyframing work (particularly for lipsynch). The face encompassed let us say 40 morph targets… Amazing…

Eventually, they show the enormous work done by animators. Because gollum was partly mocap-ed and partly keyframed based on Serkis’ performance during the takes. And the animators have added the details andy should not do because of his physical constraints.

I bow… These animators are genius.

Han, I got a bit OT :blush:

#59

Seriously though, they are amazing!

#60

My turn :slight_smile:

Ok, I tend to get lost easily when box-modeling. This is primarily because I find it very hard to add detail. Especially to add edges to an existing mesh in order to “draw” an edge loop where I’d like it; a tendon for example. What makes this difficult for me is knowing where to cut through the existing mesh, so as not to create any polys with more than four sides, or where to end a loop without creating poles with more than five edges. Especially if the loop I am trying to draw is particularly curvaceous when compared to the surrounding mesh.

Spline modeling seems to “Click” with me. I don’t have any traditional, or indeed any artistic background at all. But I find that when I look at just about any object, be it organic or solid, I tend to imagine curves flowing around its general form. For me, drawing splines and patching them seems totally natural. However, the same problem arises later with detailing. Where and how to add edges without totally disrupting the existing mesh.

Are there any rules of thumb for adding detail? For instance, I am aware that if cut across the corner of a quad, you get a tri and a pentagon, which can be remedied by adding an edge from a corner of the pentagon to the middle of the triangle.

How about for beginning and ending new loops. Should you generally start on an edge, or a vertex? End on an edge or a vertex? I seem to gravitate to vertexes and I think this causes me a number of problems with unwanted triangles, or six edged poles.

Most tutorials I have found paradoxically abandon the reader at the detailing stage after a great deal of hand holding at the more intuitive general layout stage.

If anyone has any tips to offer when it comes to detailing, I’d really appreciate hearing them.

Gelfie