Comparing topology of old and new models

[jujube]

I unzipped some of the old downloads to get to some of the ancient iterations of base.obj. I'm uploading the blend file for the convenience of anyone else who might want to play with them.

The most obvious thing is that all the alpha meshes are in the T-pose. ***The difference between alpha6 and alpha7 topology is surprisingly dramatic. They got rid of the clavicle! It seems they put back some of the definition in 1.0, but it doesn't have those lovely neck muscles that the old alphas apparently had.*** Also, for some reason I prefer the alpha1 eyes when seen from below.

(And while the "Professional mesh topology" wiki page says that only the current version is all-quad, it turns that so is the alpha6 mesh; only its joint helpers are quads. So it's not clear why the artists got rid of those lovely neck muscle loops. )

Lindsay, what are your thoughts on this? There definitely seems to be more shoulder muscle definition in the pre-alpha7 models.

edit: ***pretend I crossed out those three sentences; the topology is the same from alpha6 to alpha7, I just forgot to hide the tights helper when I took the screenshot of the alpha7. Also, I am a complete idiot.

[blindsaypatten]

It is a shame the neck tendons got attenuated, they are often cited as an important component of realism. The Alpha 1 ear is also way better than the current ear, it would be really interesting to have heard the conversation that led to that change.

I don't think the designer was very concerned with realistic musculature, I think they just wanted to get the general build with an expectation that either characters would have clothing on or that realism wasn't critical. Of the people posting on the forum I think we two are the only ones with much interest in realistic musculature.

It's pretty clear that the base mesh is a female character. I can remember wondering why the base mesh was different than the half and half startup mesh, and I can remember suddenly realizing why, but I can't remember what the reason was right now...

If you flip back and forth between the Alpha 1 mesh in a t-pose and the current mesh, you will notice that the modeler didn't account for changes that would occur to the upper-shoulder/lower neck area, they just bent the arms. In both the way the shoulder and chest muscles are merged isn't at all realistic.

I have a love-hate relationship with MakeHuman. This is largely because it is so promising with the way it makes it so easy to produce a plausible looking human, but then makes it so hard and complicated to produce something specific. One moment you are gleefully enjoying the simplicity of pushing sliders and the next you have to be a Blender guru just to achieve something that seems like it ought to be the default, like glossy eyes or non-glossy skin. I can understand that MakeHuman is MakeHuman and Blender is Blender and I'm taking the output from MakeHuman to Blender so it's my responsibility to handle working in Blender but... Couldn't there be a slider in MakeHuman for skin glossiness and one for eye glossiness, and then just have them be that glossy when imported to Blender, without ever resorting to a node editor?

The analogy is a bit of a stretch but using MakeHuman is like driving a car, and using Blender is like being an engine mechanic. Cars would be a lot less popular if everyone had to be their own mechanic.

I'm whining again so I'll shut up now. All I have to do is to think back to making human figures from scratch to see the value in MakeHuman.

[jujube]

The old topology basically has all the things I've been wanting, even my nostrils and ear holes! And I was wondering why base.obj has those sunken shoulders, so that mystery is solved. The alpha7 even has better topology on the tights helper - the current tights helper directly follows the topology of the human mesh, while the alpha7 helper shirt is made of nice neat evenly spaced squares. Honestly, I'm starting to wonder if it wouldn't be better just to return to the alpha mesh. It also has the T-pose which seems to be more standard in character creation.

Yesterday I was playing around with UV unwrapping the head to fit the alpha7 head to the current textures. (I turned the head into a clothing item and it looks fine, except for the lack of texture.) I got one ear lined up with the texture, and it looks great:

ear.PNG (99.22 KiB) Viewed 7906 times

In another thread punkduck mentioned the conflicting user needs for high vs low poly meshes. Which is basically the purpose of the proxy feature. But I think it would help the end user for the default, low-level mesh to optimize its loop flow to best reflect human anatomy. (Which would make it a better base for advanced sculpting, but also make it easier for the casual user to make a mesh that looks good, with minimal tweaking. )

Oddly, although the current iteration of the mesh was made by Manuel, in his human-making blender addon he seems to have put the neck tendon loops back in.

The issue of needing to correct the materials in blender is less a limitation of makehuman and more about the limits of whatever file format you use to export. I feel like it should be possible to modify the MHX2 i/o to allow custom glossiness levels. (Sadly, I know nothing about writing blender addons. ) Before I discovered/learned how to use MHX2, I had to manually select the cornea and change it to a transparent material every time I loaded a new human or else the eyes would be a creepy blank void. Which was a a huge pain.

ManuelLab was created as a blender addon to make it more friendly for the end-user, I think. But even though most of us use blender, I still like that MH offers the freedom to export to whatever program you want, since it's a standalone application. A fully functioning perspective camera would make things easier too, I think.

The purpose of makehuman was to be only a character creator, where you would do the actual artwork and posing in a different program. So the sliders are perhaps deceptively simple. But to me it still has a huge amount of potential. I'd love it if someday it would be possible add a user defined rig+mesh as the basis, in makehuman and in maketarget. It's supposed to be MakeHuman, not MakeAnimal, but I think adding new rigs and models is where makehuman has the most potential. If we turned it into a one stop shop, with the ability to create full scenes in makehuman, or have perfect photorealism, there's no way it could compete with bigger programs like Daz. The simplicity and moddability is where its true potential shines.

Since makehuman is open source, this means that all the developers are volunteers. So basically, you and me have hired ourselves as devs!
I think we should make a wishlist of all our modeling/art asset to-dos, making it easier to chip away at these things (for us, or for whoever else wants to work on it. )

[brkurt]

I often start with the preAlpha 5.1 mesh, as the shoulder design suits my purposes better. I normally add a loop cut between the upper arm and shoulder girdle, so that arm moves more freely from the shoulder. Here is an example:

http://www.makehumancommunity.org/forum/viewtopic.php?f=14&t=13935&p=36947&hilit=perfect+punch#p36947

[joepal]

The old topology basically has all the things I've been wanting, even my nostrils and ear holes! And I was wondering why base.obj has those sunken shoulders, so that mystery is solved. The alpha7 even has better topology on the tights helper - the current tights helper directly follows the topology of the human mesh, while the alpha7 helper shirt is made of nice neat evenly spaced squares. Honestly, I'm starting to wonder if it wouldn't be better just to return to the alpha mesh. It also has the T-pose which seems to be more standard in character creation.

If an older base mesh fits your particular use case better, then why not simply make a proxy of it and use that?

Ok, moving a T-pose mesh to the current A-pose make take some tinkering, but still?

If I remember it right, the main motivations for redoing the mesh back then was that the old mesh had an uneven and very inefficient distribution of mesh density, and that several loops were made in such a way that rig deformation and weight painting became tricky.

Contrary to popular opinion, it wasn't a whimsical "hey, let's wreck the whole system cause I like piss users off" decision. It was the end of a process with up to a decade of discussions, experimentation and user feedback.

In another thread punkduck mentioned the conflicting user needs for high vs low poly meshes. Which is basically the purpose of the proxy feature. But I think it would help the end user for the default, low-level mesh to optimize its loop flow to best reflect human anatomy. (Which would make it a better base for advanced sculpting, but also make it easier for the casual user to make a mesh that looks good, with minimal tweaking. )

High resolution and defined body features are at one end of the spectrum, where low poly and a tight mesh economy is at the other. It's tricky to accomodate both in a single mesh. And the base mesh doesn't intend to do so. It's a vanilla least common denominator. For specific needs it is expected that it is combined with a proxy. Or that steps such as sculpting is taken outside MH.

Since makehuman is open source, this means that all the developers are volunteers. So basically, you and me have hired ourselves as devs!
I think we should make a wishlist of all our modeling/art asset to-dos, making it easier to chip away at these things (for us, or for whoever else wants to work on it. )

Yes, the basic workflow here is that if someone wants something, then the first suspect for implementing said feature is that very person.

There's nothing magical involved in the procedure, no big hoops or procedures that need to be observed. If you think something should be done, then do it. And when something is contributed, we can discuss how to best fit that into the whole.

The current crew mainly consist of programmers, and our current development focus is to bring the core MH code up to date (it is currently stuck with code dependencies which are reaching end of life soon, or which have already done so).

You can make a wish list with art issues if you wish. But currently it is unlikely any of the programmer devs would suddenly become focused on art and deal with such issues.

[jujube]

I'll probably respond to the rest later but basically, I agree with you.

But the "wish list" was just meant to be a to-do list for me to grant my own wishes (or perhaps for any of the other users who may want to work on them). I understand that there are no artists on the list of "official" devs*. Thinking about going through Lindsay's posts and making a list of all the suggestions, for my own reference.

I'm working on the proxy thing right now. I haven't figured out how/if you can make a proxy based on a posed mesh, but I'm working on making a "proxy" of just the ears. None of what I said was intended as a demand on the current team of programmers, it was intended entirely as something that me and the other forum artists would work on. Sorry if it came across as overly critical of the team... for me the frustration is something that motivates me to turn makehuman into what I want it to be, with the help of these discussions.

(Although, about the mesh density, the current mesh isn't free of uneven density either. But I don't know anything about weighting or rigging. )

But when I have the time I also want to try my hand at documenting some of the makehuman code. Then it will hopefully be easier to have more of the modding that I dream of.

*official in quotes since anyone who contributes to the functionality of makehuman is effectively a volunteer dev

[blindsaypatten]

(Although, about the mesh density, the current mesh isn't free of uneven density either. But I don't know anything about weighting or rigging. )

It is interesting how many faces were devoted to the belly button on Alpha 1, there is also more than a simple grid between the legs.

The ears are an interesting contrast, the alpha mesh looks like someone took an actual ear topology and tried to fit a mesh to it, the current mesh is clearly a very regular extrusion grid which has been smushed to produce something ear-like.

[jujube]

My first attempt at creating an alpha6 proxy:

Top image is base.obj, bottom is the 33% startup mesh.
For the longest time I assumed you couldn't make clothes on a posed/targets-applied mesh; I'm glad I proved myself wrong. It's just the head that's messed up. Which is weird, since the head by itself works as clothes:

The procedure I used:
1) create "basemesh with helpers" via makeclothes
2) with human selected, import default.mhskel (I wanted to link to the rig i/o addon here but I don't remember how to find it)
3) pose human to match alpha mesh
4) on human:
i) clear parent and apply armature modifier
ii) delete all the vertex groups except for mid/left right
5) make clothes