PDA

View Full Version : I want to model different faces with same topology.


animatics
07-31-2009, 11:15 PM
Questions i have:
- Tips to take pictures to people for 3D modeling. Websites with good pictures for 3D modeling.
- Is the model's topology good, and is it usuable for different subjects?
- Is the technique i want to use to spare modeling from scratch good? Are there better ones?

Sometimes is hard to find good references to model and study modeling

For the first time me, my collegues and teacher are having a shooting session using 2 cameras with high Focal Distance values (100mm-300mm to kill perspectitve) to capture our faces and eventualy one of us doing full body poses.

After the shooting the intention is to each person model it self and the poses.
Now... Each person is a person, like each model. But what if you get a topology, that you might think it fits well generally (more towards animation) and you would like to apply it to different faces?

Idea: Use one model with a topology at taste and adapt it to each of my collegues face. So i get their shape but i don't need to do a model from scratch and think in topology again and again.

About the Model: Used only minimum Key and Fill Loops to get a workable shape. Ears and Nose already with some detail so i don't have to bother with that each time.
The colors show where i started to model from: the "holes" (Eyes, Mouth, Nose, Ears) and then connected those circular loops with 5 edge poles. The only Poles that are not from topology are the ones over the eyes(Edge Reduction to the back of the head)

Technique:In Lightwave's Layout use Joint bones to control the shape of the mesh in different areas, with an image in the background. In the end, Save Endomorph.
The obcjective is to get a good shape to start and add Loops for detail, not a finished model.

Attachments:
-> (http://www.mediafire.com/?nnxmtfmwzyl)Video File: (http://www.mediafire.com/?nnxmtfmwzyl) Fast slopy prview of what i want to do with photos.
-> (http://www.mediafire.com/?sharekey=c8847043c48603f2d1014a7a667fa2b42372ba0c139dadffc95965eaa7bc68bc)Model and Scene files: (http://www.mediafire.com/?sharekey=c8847043c48603f2d1014a7a667fa2b42372ba0c139dadffc95965eaa7bc68bc) The model's shape is hideous, but that's because it has no goal shape yet.

Cheers

http://i153.photobucket.com/albums/s202/animatics/Lightwave/model.png

Animasta
07-31-2009, 11:21 PM
edit: forget what I wrote..

animatics
08-01-2009, 03:51 AM
What did you wrote? :D
Ok, ok, nevermind...

cowtrix
08-01-2009, 07:12 AM
I've seen this before, the challenging goal of an all-encompassing topology. A decent test it can you model an extremely thin, malnourished person and an obese person with the same mesh. Repeat for female and male facial proportions. If those four models look great, then your topolgy is all encompassing.

I believe this may be the wrong forum for this thread though. Organic modelling would be your best bet.

animatics
08-01-2009, 11:36 AM
I don't have colleagues like that ehehe, but it sound like a good test =) I could give it a try with a Skull ( We have a closet full of bones hehe) and some photoshop obese person.

Yes, i didn't noticed the General Techniques>Modeling Forum. If an admin can change this thread there, it's good.

Cheers

mister3d
08-01-2009, 11:57 AM
There are some interesting tips there. http://www.subdivisionmodeling.com/forums/showthread.php?t=8911

animatics
08-01-2009, 12:41 PM
That's a great Collection Mister3D.

Thanks

Laa-Yosh
08-01-2009, 11:15 PM
I don't believe in generalized topology. There are some similarities but every single head I've modelled had pretty different loops in the end.

mister3d
08-01-2009, 11:42 PM
I don't believe in generalized topology. There are some similarities but every single head I've modelled had pretty different loops in the end.

I agree. Nowadays with zbrush it doesn't make sense to keep any kind of topology templates, as you start in zbrush, and then just build over.

MasonDoran
08-02-2009, 12:11 AM
It becomes a concern when you have to optimize a pipeline to fit a budget. Say you need 20 characters. with very tight constraints. Employers put a lot of pressure on getting things done quickly and that means cutting corners where you can without losing quality.

If all the heads had identical topology (vertex ID) then creating blendshapes as different characters will speed things up. This will also speed up rigging the character. UVs can also be inherited, and even if you get stretching that is very easily fixed, as opposed to re-creating everything from scratch.

Michael5188
08-02-2009, 12:12 AM
All the extras in Pixar films originate from the same topology, they look extremely varied. (not style wise but body type/appearance) If done well it can be a huge time saver.

animatics
08-02-2009, 06:13 AM
I agree. Nowadays with zbrush it doesn't make sense to keep any kind of topology templates, as you start in zbrush, and then just build over.

Zbrush? This is thought towards animation not sculpting.
You do your topology in Zbrush? Well i would like to see that in action. Is it handy?

Imagine i take like 8 photos from my colleagues, and do 8 models from scratch? Of course, like it was said, every face has it's own topology, but they all have a lot of common stiching points in all those different topologies right?


Here's an example of a couple of minutes spent adding somes loops and dragging points. No reference here since i don't have the photo yet.

http://www.newtek.com/forums/attachment.php?attachmentid=75889&d=1249172762

After looking to you Collection Mister3d i think i'll change something in the nose, in the chin and maybe the mouth.

Cheers

Laa-Yosh
08-02-2009, 12:08 PM
It becomes a concern when you have to optimize a pipeline to fit a budget.

Fortunately I don't get jobs like that, we're usually expected to deliver quality instead of quantity.

earlyworm
08-02-2009, 12:50 PM
For photoreal human characters, you can quite easily get away with using a single base mesh. I've got a bust of a male head at work that I use for reference, I'm able to reuse that mesh without any problems or reduction in quality for different male characters.

I'm sure there are situations where I would need to modify it to take into account different faces, but they are few and far between.

ThirdEye
08-02-2009, 01:04 PM
I don't believe in generalized topology. There are some similarities but every single head I've modelled had pretty different loops in the end.

Yes and no. There are some things that are always 90% similar in all the human heads you can look at. Ears are ears, eyes are eyes, mouths are mouths. They're often very different, but usually not enough to justify a totally different topology. If you're talking about creatures and not humans then it's another story.

animatics
08-02-2009, 01:30 PM
Fortunately I don't get jobs like that, we're usually expected to deliver quality instead of quantity.

So, Modeling only good in Zbrush? Well not everyone uses Zbrush, and i did not see any examples of modeling than doing topology in Zbrush yet.
Don't you repeat many aspects of topology from model to model?

Well if you model very different creatures that's another thing.

I think there's no loss of quality if the base model is good and you know how to use it (not saying my base model is good or that i know how to use this at the moment...) In fact i think it's "quality time" being in detail step faster.

Cheers

Laa-Yosh
08-02-2009, 01:32 PM
Yes and no. There are some things that are always 90% similar in all the human heads you can look at. Ears are ears, eyes are eyes, mouths are mouths. They're often very different, but usually not enough to justify a totally different topology. If you're talking about creatures and not humans then it's another story.

Unfortunately I can't show any wireframes, but trust me, every single human head we've done in the past 4 years is different, some of them drasticaly.

We've also tried the unified topology stuff once, with a lot of the detail coming from displacements, and the only visible difference in the resulting movie was that some of them had a beard. Apart from that they looked the same.

If you reuse a base mesh, that's what you'll get - far too similar and mediocre looking characters.


And it's not just the various shapes of facial features, their placement and size and such matter as well. Not to mention age differences which can completely change a face.

Laa-Yosh
08-02-2009, 01:50 PM
So, Modeling only good in Zbrush? Well not everyone uses Zbrush, and i did not see any examples of modeling than doing topology in Zbrush yet.

We use Zbrush to create concept sculpts for the faces, and then use either Topogun or 3ds max to build a high res model on top of that. We go for 99% similarity and try to leave only skin detail to the displacement maps so that we have a good level of control in the facial animation.
See here:
http://zbrushcentral.com/zbc/showthread.php?t=073117

Don't you repeat many aspects of topology from model to model?

Some patterns are there on similar faces, but the shapes of eyelids, noses, ears, brow wrinkles etc. are usually very different. If you aim for realistic and characteristic results then you will create very different forms, which require different topologies as well.

Well if you model very different creatures that's another thing.

I've built about a dozen high res human heads in the past 2-3 years, that's what my experiences are based on.

Ruramuq
08-02-2009, 01:58 PM
We should not confuse sculpting with modeling.

there are rules that justify topology of every organic creature, very complex or very simple rules, but essentialy the same rules.(but most people dont see that, thanks to internet)
using zbrush is skipping those rules, for a benefit perhaps of speed or visuals but not for animation quality.

nobody has tried to implement something complex yet. so we can't discard future possibilities. that wouldn't be fair.

Laa-Yosh
08-02-2009, 02:08 PM
Huh?

The idea is to separate the artistic and technical issues of the task. First you take something even as simple as a box and create a detailed sculpt from that, without any restrictions or considerations.

Then you take the sculpt and lay out the edgeflow on top of that, making sure that you follow the forms and the expected direction of facial (or any, as this can be applied to anything) deformations.

animatics
08-02-2009, 02:13 PM
Thanks for your input Laa-Yosh, still i won't discard this possiblity =)

Ruramuq
08-02-2009, 02:32 PM
The idea is to separate the artistic and technical issues of the task. First you take something even as simple as a box and create a detailed sculpt from that, without any restrictions or considerations.That method is artistic not really technical, sculpting first then finding the best way to make it work.
Topology is more independient than a few particular models. something that you can apply to other models
I see different layers, not one final sculpture that needs a topolog,y because Sculpture can be extremely delailed, the topology in those cases must be extremely detailed too, but thats not the case, not at all. so the discussion of topology related to sculptures is usually very unrelated from my point of view

Laa-Yosh
08-02-2009, 02:34 PM
Can't follow you, sorry...

GrogMcGee
08-02-2009, 05:35 PM
Laa-Yosh, you've talked mentioned that you (and your team?) don't use generalized topology - and that you can't show any pics... ah well, but perhaps you could do some over draws of what the kind of topology you'd use on some photos of real faces. I know that people would find it useful ... :shrug:

ZacD
08-02-2009, 05:42 PM
When I think of generalized topology I always thing of those video games with those
"face and body" editors, create a person or whatever, and those tend to yield decent results and can look very similar to real people.

Laa-Yosh
08-02-2009, 05:48 PM
That's not a bad idea, might try it... however I usually do some reworking of the loops here and there during the modeling process.

Which is also why I don't spend any time on trying to pre-draw the topology on the sculpt, as seen in some tutorials... it actually takes longer to erase and redraw lines then it takes to reorganize the geometry.

Most of the stuff I've seen here on this previous link (http://www.subdivisionmodeling.com/forums/showthread.php?t=8911) is pretty good though.
Funny to see a few old screenshots of mine too ;) even if now I see a lot of problems with them...

Anyway, the best source I know is still this one from Gollum.
http://express.howstuffworks.com/gif/gollum-7.jpg

musashidan
08-04-2009, 12:01 PM
That method is artistic not really technical, sculpting first then finding the best way to make it work.
Topology is more independient than a few particular models. something that you can apply to other models
I see different layers, not one final sculpture that needs a topolog,y because Sculpture can be extremely delailed, the topology in those cases must be extremely detailed too, but thats not the case, not at all. so the discussion of topology related to sculptures is usually very unrelated from my point of view

Retopo-ing over a detailed sculpt is probably the greatest advancement in sub-D modeling since it's inception.
The super fine detail can obviously be mapped.But retoping allows for extremely fast and accurate all quad meshs with poles planned out well in advance rather than re-routing on the fly.i personally find it to be more efficient,fluent,artistic,natural and abve all less time consuming.Of course this argument will always be composed of those from both perspectives of artistic-Vs-technical.

Plus:Tamas Varga is a sub-D king :D

Ruramuq
08-04-2009, 04:04 PM
Retopo-ing over a detailed sculpt is probably the greatest advancement in sub-D modeling since it's inception.
Yes, but I was not talking about Retopo thing, thats another matter.
I don't know what you mean, but I was using the title of this thread "different faces with same topology. "
again modeling should not be mixed with sculpting if you are pursuing a better technique. but there's a lot of mixing, repeatead over and over, from the same person.

in difference to you opinion, I completely disagree about Laa-Yosh experience in modeling.Completely.
Don't wanna fight, thats why I just posted my opinion previously, and nothing more about him, and his explanations.

not sure, but I think I already said in topology forum, because one model becomes famous like the golum one, does not mean it is a rule.. wise people should be careful about that. how open is your mind to new ideas. that's what gives experience IMO.

now its not hard to figure it out why many complex ideas stay private, and the more popularized are the cheaper

and excuse me if I sound annoyed. but the same topic about mixing zbrush with everything is becoming very ridiculous to me.
if a person has a method that works for whatever he, and only he does, good for him, in its individuality and his will to share it, NOT to impose it as a fact for the rest of us, or the naive ones.

Laa-Yosh
08-04-2009, 04:32 PM
Seems like I'll have to do some tutorial to make my statements clear...

Laa-Yosh
08-04-2009, 05:00 PM
because one model becomes famous like the golum one, does not mean it is a rule..

What you still completely fail to understand is that it's not Gollum's exact topology that people should look into, but how it's all guided by the underlying forms and the deformations it has to accomodate. That's what - IMHO - people should learn and apply to each different model they want to create.

animatics
08-05-2009, 02:29 PM
Only today i noticed this one
http://forums.cgsociety.org/showthread.php?f=25&t=38469 :arteest:
Quite similar and very devoloped =)

Still, i would like to see that tutorial, Tamas Varga =)

I understand Ruramuq point of view, in a way that many ppl dont use Zbrush to make the models from scratch. So inputing Zbrush closes the possibilities for many ppl in this matter. Still if it works really well is great to know about it =)

Thanks guys

GrogMcGee
08-05-2009, 03:00 PM
Only today i noticed this one
http://forums.cgsociety.org/showthread.php?f=25&t=38469 :arteest:
Quite similar and very devoloped =)

If you skip to near the end of that thread you'll see that the same points are have been brought up.

What tends to seem to get lost is the impact of Mudbox and Zbrush - with an application like zbrush why would I ever bother to make the mesh in max or maya or whatever? I can make the whole character in zbrush then create the a loRes in whatever application I feel like.

The time invested in sculpt -> retopo -> generate maps; the retopo part is negligible - now the important point here is that it's damn important to know how to create a loRes mesh quickly, which means we should continue to build those meshes in or favorite component modeler just as much as we need to sculpt.

I guess my point is that sure most of the time you might be able reuse a base mesh, but what about those time when you find that the silhouette is just wrong or worse that mesh just doesn't support the detail you need to put on (or really bad both). The likely cost of trying to come up with a nearly perfect topology is surely outweighed by the ease of just doing the retopo ...


Side note:
What's up with Topogun? Are they dead? Last update (to the website) was in '07 ... is it effectively abandon-ware now?

Still, i would like to see that tutorial, Tamas Varga =)

I'll second that :bounce:

Laa-Yosh
08-11-2009, 02:24 PM
Since I've been working on this toady and it doesn't really belong to any of our projects, I thought I can show it... This is not a tutorial on facial topology though.

****
Short explanation though... so, I've said that I don't believe in generic topology and I stand by it; however, I did a little testing for a few projects now on speeding up the blendshape creation workflow. For this I've built a sort of a 'face mask' mesh that's pretty low res. I build this on top of an existing face, in this case the main villain of the AC2 cinematic, then I use a wrap deformer to transfer all the main blendshapes of the character to the face mask. So the result is a blendshape library for the face mask.

Then in a new scene I create an additional blendshape for the face mask and I fit that to a new character. I wrap the new face on top of this face mask and use its blendshapes to transfer the deformations to the new face.

So it's like this:

old face -> face mask -> new face

The quality of the deformations obviously will suffer a lot, so there's still a lot of work to do. Give characteristic features to the new shapes, add wrinkles, maintain volume etc. etc. but it still saves us a few days.
****

Anyway, the face mask is somewhat representative of how the main loops of the face could be layed out. It does not have even spacing, wrinkles, or any facial features really, but it's fine to show the general directions of facial deformations. I've put together this little image to illustrate what I mean. I've marked the typical wrinkles in blue and also where the skin compresses a bit in red (note how these tend to overlap!), whereas green shows where stretching occurs.

Here's also what blendshapes are combined:
0. nothing, obviously ;)
1. jawOpen + browsUp
2. browSqueeze + smile + squint
3. browsDown + sneer + pucker + cheekPuff
4. upperLipUp + lowerLipDn + frown + topLidUp + bottomLidDn + dimpler

So if you build a face, this is one of the guidelines you should follow: the general directions of facial deformations.
The other guideline should be the exact shapes and proportions of your face, and the pattern of the wrinkles (which is usually almost as different and specific to each person as fingerprints are).

Hope this helps, and we'll see about that topology tut thing later...

animatics
08-13-2009, 01:06 AM
Funny
What i was trying to do for modeling you are kinda trying to do it for animation. =P

Thanks a lot. This is actually what i was looking: a good base mesh for deformation and understanding how the topology should/could be to have a better performance.

I'll be sure to study these images for a while =)

Cheers

Laa-Yosh
08-13-2009, 10:21 AM
I have to re-emphasize this:

The mesh you see is only good for deformation.

It does not accomodate any characteristic facial features at all, and it also has far too many poles and uneven poly distribution.
So it isn't even good enough for a base to expand upon by just adding a few more loops.

CGTalk Moderation
08-13-2009, 10:21 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.