MakeHuman

[fae_alba]

I was introduced to this app by a co-worker who used it in teaching animation for game engines a while ago. MakeHuman is free and allows you to create your own character by using pose sliders. Don't know if the character could be imported into A:M. I'm trying to bribe my co-worker into writing an import/export plug in for MakeHuman and A:M (she wrote a lot of the deformation code for MakeHuman a while ago). Might take copious amounts of beer to convince her, but hey, it might be a worthy cause!

[robcat2075]

You can certainly import the mesh to A:M, you'll still need to rig it however.

 

If your co-worker could get it to export MDLs with bones the mesh is attached and weighted to, that would be a useful thing.

 

I think it handles some deformations like facial animation with morph targets. Those would need to be converted into A:M style pose sliders. Other things, like how a shoulder bends would have to be converted to a smartskin.

 

None of that is undoable but the person would have to make a study of the A:M code to understand how A:M executes these concepts that have a similar outcome on-screen but who's internal workings are a bit different.

[jakerupert]

Lets face it.

None of these conections will ever be established since Am is and will stay

a niche product and these efforts will never be worthwhile.

No use to stick the head in the sand .

Am is still great as it is but users tend to set too much high hope in it.

Just keep it as it is and be realistic.

[fae_alba]

Lets face it.

None of these conections will ever be established since Am is and will stay

a niche product and these efforts will never be worthwhile.

No use to stick the head in the sand .

Am is still great as it is but users tend to set too much high hope in it.

Just keep it as it is and be realistic.

 

You misunderstand I think. I'm not looking to alter A:M, but investigate if this particular tool could be used to at the very least make modeling easier for A:M users. I'm horrible at modeling human characters, but some of the projects I am planning on need a whole bunch of them. If this app, or one like it were capable of creating stock characters with enough variety that can then be imported and rigged in A:M then it would help poor shlepps like myself out quite a bit. I did download it onto my laptop and tried briefly to create a character. The response was quite slow. This is the same machine that I use for A:M with no real speed issues at all, so I don't know if it is a viable solution for me or not. My co-worker did show me some of the work she has done with MakeHuman, and I was impressed with her results.

 

Also, while A:M may be viewed by some as a niche product, I personally don't think it is. A:M can go head to head with other animation apps out there.

[robcat2075]

If you export from MakeHuman at its lowest res that can import to a well-shaped, reasonably fast performing spline model in A:M. However, occasionally a spline will need repairing and that will lose the decaling from that area.

[Mechadelphia]

MakeHuman is pretty good.

 

I've imported a few meshes into A:M but I find that they are way too dense to be useful in A:M and the time necessary to clean them up and fix spline continuity issues is not worth the effort to me.

 

Rather than convincing someone to program a way to import and export to and from A:M I personally would go about it a little differently.

 

Here's an idea:

 

MakeHuman uses the same base mesh for both male and female characters. You could export the base mesh then import into A:M. Then, the model could be cleaned up and optimized once for the way A:M works.

 

Then, create a set of pose sliders for that model that emulate the functionality of MakeHuman for generating characters. (You could even set up some presets with a few extra sliders.) Export the mesh from an action window when you are happy with the results.

 

I suppose that rigging could be factored into a setup like that but that would take a little more work.

 

What do you think about an approach like that?

[fae_alba]

MakeHuman is pretty good.

 

I've imported a few meshes into A:M but I find that they are way too dense to be useful in A:M and the time necessary to clean them up and fix spline continuity issues is not worth the effort to me.

 

Rather than convincing someone to program a way to import and export to and from A:M I personally would go about it a little differently.

 

Here's an idea:

 

MakeHuman uses the same base mesh for both male and female characters. You could export the base mesh then import into A:M. Then, the model could be cleaned up and optimized once for the way A:M works.

 

Then, create a set of pose sliders for that model that emulate the functionality of MakeHuman for generating characters. (You could even set up some presets with a few extra sliders.) Export the mesh from an action window when you are happy with the results.

 

I suppose that rigging could be factored into a setup like that but that would take a little more work.

 

What do you think about an approach like that?

 

Actually that is an alternative that I am considering. Perhaps when there is some downtime, I just might tackle it.

[Rodney]

Yes, in my quick tests... not bad at all.

 

In order to be really useful you'll need an intermediary conversion to lower the density of the mesh created with MakeHuman from 'ridiculous' to something approaching 'optimal'. The mesh can come through pretty clean but is on the order of 20 times more dense than it needs to be.

Moving the mesh through a different format on it's way to A:M (besides just OBJ) might help to lower the density of a mesh to something workable (esp. for animation). And isn't that really the goal for humanoid characters... to animate them?

 

Aside from direct usage... many tools like this can be quite useful in creating derivative products such as Decals.

Whereas it might take considerable time to setup and decal one from scratch in any given piece of software with MakeHuman (or products like Project Pinnochio) you might just create... borrow... and modify a decal for use with your own character in A:M. I'm not sure about how high rez we are talking here but... at least it could create a starting point.

[Malo]

A few years ago, I started a similar project. It was not finished, and it is smaller, the model would be also reviewed. But if it can give ideas or be useful to someone. here is the link:

Model to download

Just go into action mode, then put the "pose1" on "on". Move, rotate, scale the bones and when the proportions are correct, correct modeling by moving "CPs".

 

The original idea to create an import/export bridge between MakeHuman and AM is good.

If it is to do the final rendering in "AM" ... it will be a long and hard work for the programmer to convert poly to patches and may be impossible

If this is to use AM to animate characters imported and then export them to an external polygonal renderer, the idea is good and it is possible for a person familiar with the programming and 3D formats, without too much difficulty.

[Rodney]

Thanks Malo, that's a sharp looking fellow!

I agree that creating a modifyable character in A:M is the way to go.

I've often thought of taking a project like that on myself (perhaps using Squetchy Sam even) and... then I think of the work involved... and find something else to do.

 

Here's a few renderings of your character so that folks can see what he looks like.'

I've adjusted his skin color from blue to distinguish him a little more from Squetchy Sam.

 

Nice character!

[Rodney]

I'm sure I'm using the setup wrong but wanted to play...

 

Here are three quick modifications of his face:

[Malo]

Use "Squetchy Sam" would be a good idea. It is well built!

I also think a new feature that allow to export and import the poses would be a big plus for such a project. This would allow everyone to enrich poses deformation, as is done in Poser.

[fae_alba]

I suspect that in MakeHuman, the super dense mesh is needed to allow such a wide range of deformations (male/female, old/young, tall/short, fat/skinny) especially in the face. The real question in my mind is can the same range of changes be done using an A:M model, with a moderate number of splines? The way it was pitched to me, you can create any kind of character (humanoid at least) from this one basic model, simply by adjusting pose sliders. I'd love to give it a go, but with Rear Window as well as two other shorts that I am starting work on, and three blogs...I'm a busy bloke. Oh, just got word that I have a good chance of an excellent new job offer coming my way in a way or so, which puts me on the road 50 percent of my time.

 

Keep the conversation going, beating ideas around and someone experimenting until a solution is found is what makes A:M so great!

[Rodney]

"Squetchy Sam" would be a good idea.

 

This is pushing us farther off topic but... for the recent contest one false start I had was to begin modifying Squetchy Sam into a team of different characters.

 

Samantha was one result:

[Rodney]

The new release of MakeHuman is out.

Some very nice improvements too!

 

Several things make this release a great tool for use with A:M.

 

One of the better improvements is that geometry under the clothes you add can be removed upon saving.

This considerably simplifies the mesh.

It also separates parts of the mesh so they are easier to modify.

Then in A:M hide what you don't want to look at and work on what you do.

 

The workflow I tested:

- Export from MakeHuman to OBJ Free This is where decals get saved

- Convert from OBJ to FBX (via FBX Converter) Free This is where I insure that I do not reintroduce tris (Make Human basic mesh is all quads)

- Convert from FBX to OBJ (via FBX Converter) Free This is to get it to a format that will easily import into A:M.

 

Why convert to OBJ twice... I dunno... it seems to work better and keep decals and such.

I'm guessing I could export intially to Collada etc. but I'm not as familar with that format.

As it is this workflow... works.

 

Note that most importing of OBJ files A:M may appear to lock up upon import of the OBJ file into empty model take a break and let it run.

When you return it'll be done.

 

The OBJ importer is definitely doing its work because not only am I getting proper decaling with UVs I'm also getting Groups!

The Groups don't always get everything (such as where a jacket sleeve penetrated through... no doubt this confused the normals) but they are easy enough to edit.

 

The density is still high but certainly workable.

 

Most definitely worth adding to your toolbox.

 

http://www.makehuman.org/blog/makehuman_101_is_ready.html

 

 

Added: I should have also mentioned that there are some basic rigs you can add to the meshes. I don't think these will be particularly useful but... one is BVH and others use JSON. This might be worth looking into for conversion purposes.

[Simon Edmondson]

The new release of MakeHuman is out.

Some very nice improvements too!...

 

 

Rodney

Thank you for the update info.

I tried it for the project currently underway, using the previous version, as my modelling skills are not of the highest order.

Making the model was straightforward and easy enough in MH, but found it very dense in AM, when converted into splines ( the economy of splines was never more apparent ). Set to and removed a lot of splines and remade the model before installing the 2008 rig in it. At that point difficulties started to arise.

 

The MH model arrives with the arms down and the forearms angled forwards. It was awkward to get the rig installed in that position ( for me at least ) and, when installed, the mesh didn't behave as expected. Spent a long time weighting the cp's to the rig but, when starting to animate with it, the model didn't behave as desired. So started to use Snap to Surface to remake the model in the customary T position within AM, so I can develop modelling/spline skills further and have a workable model when the blocking is done.

 

Had hoped it was going to be quicker than that but, it was not be.

regards

simon

[Rodney]

I sympathize. Rigging is an art I have yet to master.

But that is mostly because I don't put the proper effort into it.

For me it's one of those things that because I'm lazy and think I can use shortcuts I skip steps and pay the price later for it.

 

I have a theory that the best riggers are those who when young were apt to model (plastic model kits) and put together detailed puzzles and such.

I always liked the idea of putting projects together like that but rarely had the patience for it.

 

Part 'o my problem is that I always think there surely must be a better way to do it rather than just do it.

[robcat2075]

One of the things I like about TSM2 is that you can rig models that aren't built in the standard T-pose, they could be in any arrangement.

 

That said, I think most of the other rigs can be maneuvered into other forms also. If you have a "policy" question about how to do that with a rig someone here can probably answer it.

 

 

 

Make Human is a fun idea but it makes models that are unnecessarily dense for A:M purposes. I'll note that the Transfer_AW plugin does a good job of transferring weights from a lo-res mesh to a hi-res mesh. You can rig a lo-res mesh and Transfer_AW will weight the hi-res mesh based on where its points are in relation to the lo res mesh.

 

However, it does have parameters that need to be set right for different circumstances, it is not a one-button press solution.

 

 

 

I'm not sure why people have trouble with rigging. They often say they have trouble with it but they don't say what the trouble is.

[pixelplucker]

Was a program out called Quidam from N-Sided, some of the old Amapi programmers had gone there. Program looked really promising but not sure what happened to it.

 

I personally would like to see a new auto rigger for AM that is joint based. Place joint types (knee, elbow etc) in a logical hierarchy and have it auto balance them. Old Electric Image had a joint types built in that you could quickly assign to bones or skeletons. Having some stock constraints speeds up the process and takes some of the tedious work out of it all.

2001 and 08 rigs are nice but still just bipeds.

[Rodney]

I'm not sure why people have trouble with rigging. They often say they have trouble with it but they don't say what the trouble is.

 

It helps if one can work all the way through to completion at least one time so that success can build upon success.

I believe that may be my primary issue with rigging: I've never had what would be considered 100% success with any rigging attempt.

I've always needed to tweak and improve after 'finishing' which technically is a stage I've never achieved as of yet.

*My* trouble is that I am lazy and don't want to put the required time into it.

But there has to be more to it than this.

 

So, what might help folks like me gain confidence in rigging?

I have a few theories.

 

What I think might work best would be a series of rigging lessons each of which is progressively longer and more complex.

Yet each of the lessons would pay off in that the goal of that encapsulated lesson would be met.

 

The trouble with most rigging challenges is that they require considerable time and commitment and very rarely... and this may be key... very rarely can be accomplished in one sitting. This isn't a matter of attention deficit but rather a simple need to move on to something with more immediate relevance. Modeling, rigging or animating with a goal of learning the process is not the same as having some longer term purpose or vision behind the effort... and an end product to assist with measuring success. What makes much of the rigging process even more difficult is the 'trial by error' nature of it. We place a bone and hope for the best. Smile. Frown. Adjust. Do it all over again. And as newbies to rigging we perform this mindless process rather than proceed from an understanding of why our current approach is best.

 

Part of my resistance must also be that *I think* there surely must be a better way to rig a character.

And thus I resist my own success.

[Simon Edmondson]

One of the things I like about TSM2 is that you can rig models that aren't built in the standard T-pose, they could be in any arrangement....

 

I'm not sure why people have trouble with rigging. They often say they have trouble with it but they don't say what the trouble is.

 

 

Robert

When I started with the stick figures originally I just put bones in without any constraints or balance systems, in a hierarchical chain, because I didn't know, let alone understand that such could be done. When TSM came out, I bought that and used it because it gave a simple, easily done system that could allow you to get on with the animating quickly. I moved over to the 2001 rig because that was featured in the TAOAM and could follow the Tutorial. Moved over to the 2008 rig because it was available and there was a comprehensive installation pdf.

 

I have reservations about it, because it is very complex and it often seems to do things that I don't want it to. For example, setting keys for the hip scale or head scale targets when moving a geometry bone ( even if no scale settings was changed in the translation process ). There are many aspects that remain unexplored because of the time factor involved, the intricacies of the nulls that manipulate the fingers for one., there are others. Thats all on the skeletal side but I appreciate the work that has gone into it and the advantages of using a standardised system.

 

The part of the rigging process that could drive me around the bend is the weighting and smart skinning procedures . I recently made a model using Make Human, converted it to splines, installed the 2008 rig, spent several days weighting cp's, did mirror bones and started to test it with an animation sequence only to find that the weighting was nowhere near optimal ( that is being polite ) and the whole process will have to be done again. I'm not looking forward to it and, before I get there, will set up the animation desired so that the new model can be fully tested. I saved the various iterations as the weighting was done and reached about 20, learnt a lot in the process but lost a lot of time that would have been happier spent animating.

 

My brother recently installed Virtual box on my Mac mini for me and I am contemplating going back to TSM using that and one of my old AM CD's under W7. I think the real problem is a slight lack of patience on my part and an acute sense of time passing when trying to develop something I want to do?

regards

simon

[robcat2075]

Thank you, Simon, for giving more detail.

 

The jump from stick figure to realistic human is a huge one. A huge one in any program. There is a reason that rigging polygon characters is something professional studios hire experts to work on full time and they spend many months on each character.

 

It seems trivial because MakeHuman exports rigged human bodies that other programs can use (I presume). It must be simple, it it happened with only one button press, right?

 

But that only works because they spent a lot of time (14 years!) studying the problem and studying human anatomy and programming ways to make it work... on that one universal blueprint for a shape... for one kind of character, a human within a certain limited range.

 

I've monkeyed with it a little bit... if i want to make a human with the proportions of the dwarfs in "Snow White" that doesn't seem feasible. Can I make a Cyclops? I don't think so. Can i make the fat lady I saw at the supermarket last week? MakeHuman can do heavy but i don't think it can do that sort of fat.

 

And it doesn't make aliens or dinosaurs or cars... it can't make or rig any of those.

 

None-the-less, MakeHuman does a good job within the limits it allows you.

 

 

A realistic human almost always looks bad in animation. Do you really need a realistic human one for your project?

 

Can you show the mesh you have now and show where you are having trouble?

[robcat2075]

I was introduced to this app by a co-worker who used it in teaching animation for game engines a while ago. MakeHuman is free and allows you to create your own character by using pose sliders. Don't know if the character could be imported into A:M. I'm trying to bribe my co-worker into writing an import/export plug in for MakeHuman and A:M (she wrote a lot of the deformation code for MakeHuman a while ago). Might take copious amounts of beer to convince her, but hey, it might be a worthy cause!

 

 

Paul, how did this discussion with your programming colleague go? Is she someone who could actually get this done?

[John Bigboote]

I just downloaded and played with Make Human. Not impressed- GREAT idea, but I did'nt see a use for it. Some of the 'poses' that control body parts are wonky, un-natural results. Things like, if you want to control the thickness of the legs you have to do each one separately with a slider that has no numeric value, so after you do the 1st you have to eyeball the 2nd- unless there was a 'mirror' switch that I did not see. I tried 4 or 5 times to export an OBJ to bring into A:M, each time it gave a confirmation that the file had been created- but NO file had been generated in windows explorer.

[Rodney]

unless there was a 'mirror' switch that I did not see

 

 

Yeah, there is a symmetry option that allows right/left or both sides to be adjusted.

 

Not impressed

 

But you WOULD be impressed if someone coded a plugin that did something similar for A:M.

I see it as it is... an additional tool not a solution.

It's like reference photos... they won't make your movie but they are a starting place in creating characters.

 

What they have now almost looks like their goal is moving toward scripting.

And as with most programs of this kind they could do that... run a script that created all possible combinations.

 

I'm trying to bribe my co-worker into writing an import/export plug in for MakeHuman and A:M

 

If the end product density will be the same then she should save her time.

The process of getting models from Makehuman into A:M about as straightforward as it can get.

Even taking a break for coffee I'd say from start to finish less than five minutes.

 

Edit: I do see a low rez option. I'll have to try that.

[Rodney]

Here's a low poly model taken from pretty much the basic/default makehuman model.

I say 'pretty much' because I intensionally added tshirt pants and shoes in order for makehuman to separate the meshes of those areas.

 

What is impressive about this process is how each of those areas survive the trip through conversion leaving named groups.

In A:M, all one would have to do is rename the groups and either color or apply materials to them.

 

Note that this model is the result with the low rez option selected.

It's still pretty high rez but its a lot less dense than the alternative.

 

I dunno... I'd have a hard time modeling something like this in under 5 minutes.

That makes 'MakeHuman' at least a little bit useful.

If we had a library full of highly detailed humanoid models ready to grab from I might think otherwise.

Model1.mdl

[Rodney]

Well, this is gonna get old fast.

 

 

If these would come into A:M at about half the density that'd be sweet.

 

Attached is Suitman...

 

Of note... some of the mesh areas, like the guy's tie, are separate but don't have named groups.

Of course since they are separate meshes they are easy enough to group.

 

I'm going to give credit for the named groups coming in properly to the A:M OBJ plugin, although the process has to work all the way through the conversion pipeline or else it wouldn't work.

 

Added: Interestingly, now that I'm looking for it I can no longer find the location to set the density of the mesh. Very odd.

suitman.mdl

[John Bigboote]

I grabbed your file, Rodney. No- I don't think that mesh is too dense at all. I guess there IS some value to this utility as this is a well-made model. I embellished on your work... added A:M Hair, SSSskin, A:MV18's screen-spaced AO... and a pinup T-shirt for good measure!

MakeHuman_HAIR.mat

MakeHuman_Rodney.mdl

[Rodney]

Not bad at all Matt!

 

I note that there is a proxy mode/topology that brings in a very low density mesh (glove hands etc.).

I think turning off the clothes option was what turned that density option back on.

I couldn't find it for a long time.

 

At any rate... here's proxy man:

 

 

Oh, and there is some scripting capability in Makehuman... I won't be able to explore that any time soon though.

proxyman.mdl

[John Bigboote]

I've been LOVING A:M's hair feature lately... it's like playing with a toy. Thinking about starting a 'digital salon'...

[Rodney]

Thinking about starting a 'digital salon'...

 

Please do!

 

Oh and BTW... not that this is any news to anyone... Makehuman can never approach YOUR humans.

[John Bigboote]

Whoever downloads my hair-groomed model(above, probably you Rodney) I'd be curious to know if it opens with all the grooming intact.

[robcat2075]

I'll note that the snap-to-CP plugin (different than snap to surface) is was added to help you replicate morph targets imported from polygon models. Perhaps useful for retaining the face rigging that was done in the other app.

[Rodney]

Whoever downloads my hair-groomed model(above, probably you Rodney) I'd be curious to know if it opens with all the grooming intact.

 

Good news. It comes through with grooming intact.

(note that mine is a quick screenshot... not final render)

[John Bigboote]

That is good news, allows for people to do 'styles' on other peoples models... or maybe even offer scalps with 'do's already prepped. Latest model has hair for eyelashes too, which I've never tried before...

MakeHuman_HAIR.mat

MakeHuman_Rodney.mdl

[Shelton]

Matt we have been talking about this in our weekly google plus meetings. I am very interested

[John Bigboote]

Matt we have been talking about this in our weekly google plus meetings. I am very interested

 

Steve- I apologize... I noticed JUST TODAY that my Hash Forum mailbox had a notification in it... it was from you- from February... asking about Hair. I'll be happy to help as time permits.

[robcat2075]

I was looking at Rodney's Model1.mdl and found an odd behavior. I can draw a selection around some CPs but not all the CPs get selected and some CPs I didn't draw the selection around do get selected.

 

Here's a screencam of it. I draw selection around two rings and i drag the group to show not all the right CPs got selected.

 

clip3547SelectionTrouble.mov

[Simon Edmondson]

I was looking at Rodney's Model1.mdl and found an odd behavior. I can draw a selection around some CPs but not all the CPs get selected and some CPs I didn't draw the selection around do get selected.

 

Here's a screencam of it. I draw selection around two rings and i drag the group to show not all the right CPs got selected.

 

clip3547SelectionTrouble.mov

 

 

Robert

I haven't tried the new version of MH yet but, did use the previous version, as mentioned before. One thing I did notice with that was that, when editing it later to remove some of the excess CP's, the continuity of spline was not as expected. The way I did it was to remove half the mesh then install 2008 up to the mirror bones point, then attend to the assignments and weighting as recommended, starting by removing the unwanted cp's. I used the comma key to select all on the same spline then deleted the unwanted.

 

Sometimes it would select the required points , eg on the spline running around the leg. Other times it would select points going around the leg but up the torso and down the arm. The ones on the fingers seemed particularly prone to that as I recall. My guess is that would be part of the conversion process ?

regards

simon

[robcat2075]

My guess is that would be part of the conversion process ?

 

Yes. Polygon models are a set of straight edges that each connect a pair of "vertices" and they don't have much knowledge of each other beyond that. Several edges may appear to be a continuous line or a ring but the polygon model doesn't regard them as being related.

 

When A:M converts polygons to patches it makes a guess about what edges to connect to make a continuous spline.

 

Polygon modelers have something called "edge loops" where the user manually selects a set of edges and tells the model that those are indeed a continuous set, but in A:M we build continuous spines at the outset and don't need to tell them that all the spline segments are part of the same spline.

[John Bigboote]

I can confirm Simon's selection weirdness- mine did same thing. It almost seems like a flaw in the A:M selection tool... I notice too some inconsistent spline bias's... where a CP out of the blue is for some reason peaked- you can see one in this render under his left side chin, the SSS makes it go 'dark'. I do LOVE what the SSS does for the fingers/hands.

MakeHuman.mov

[robcat2075]

the problem of selecting a spline with the comma key on a polygon imported model... to fix that you'd really have to get the OBJ importer to make more intelligent decisions about how to connect all the polygon edges.

 

I believe the comma tool is making the correct selection based on how the spline is connected. It's not the spline we imagine but it is the spline the OBJ importer made.

 

 

The problem with the lasso selection doesn't seem to be new in v18. I tried the same model in v15 and got the same result with some Cps out of the selection getting selected and some Cps inside not getting selected. It must have something to do with the model, I've never seen that error before.

[Rodney]

I believe the comma tool is making the correct selection based on how the spline is connected. It's not the spline we imagine but it is the spline the OBJ importer made.

 

I have a theory regarding this edge loop/spline continuity thing and at its core is the thought that A:M likely creates better OBJ files than most native polygon modelers.

Of course a talented modeler in any program will understand that continuity is important and strive for that, hence the reason for such constructs as edge loops in the first place.

The same can be said of spline models as well in that we don't always make the best choices when laying out our splines.

 

An area of interest to me is where and how the OBJ meshes deal with five point holes. There seems to be some improvement over approaches of the past and I assume it may have something to do with subdiv approximation. I'm thinking that those areas that connect at 5 points in an OBJ file could be targeted for automatic conversion to 5 point patches... but on the other hand they seem to render okay as splines. I do wonder if those are the areas that are causing slight discoloration with SSS etc. The same result can generally be anticipated for decals/textures applied. Even though those areas are quads they have to be treated specially because they aren't continuous and in comparison to the surrounding patches/quads they aren't uniformly sized.

 

I haven't tested to see if adding one more spline segment to create a dangling spline at that center connecting point would make a difference. I suspect it won't because the other four splines aren't likely continuous either. I'll have to look again but for the moment I assume that all 5 dead end/terminate at the intersection of all five.

 

 

Edit: I believe I answered my own question in that changing the 5 point intersections to 5 pointers doesn't help. Especially because removing them generally leaves areas that need to be further refined. Leaving dangling splines doesn't appear to particular help and the dangling CP needs to be about 1 pixel above the connected CPs to prevent generating new flaws in the surface. Thus its probably best to leave those areas unless in an area that is highly noticeable.

[fae_alba]

I was introduced to this app by a co-worker who used it in teaching animation for game engines a while ago. MakeHuman is free and allows you to create your own character by using pose sliders. Don't know if the character could be imported into A:M. I'm trying to bribe my co-worker into writing an import/export plug in for MakeHuman and A:M (she wrote a lot of the deformation code for MakeHuman a while ago). Might take copious amounts of beer to convince her, but hey, it might be a worthy cause!

 

 

Paul, how did this discussion with your programming colleague go? Is she someone who could actually get this done?

 

I'll have to revisit the subject with her. I left the company I was working for shortly after this, so I sort of dropped the idea. But if there is interest here, then I might be able to get something going.

[fae_alba]

I've been LOVING A:M's hair feature lately... it's like playing with a toy. Thinking about starting a 'digital salon'...

 

do you accept reservations from bears?

[Malo]

hello,
There are a few times I had fun converting a model to make human model in AM for testing.
It has only closed splines. It is textured, it has the rig. But it is not weigted.
If it interests anyone, here it is: MakehumanNudeGirlMDL.zip (9Mo)

[nemyax]

lower the density of the mesh created with MakeHuman from 'ridiculous' to something approaching 'optimal'

How much would you say is optimal?

[Rodney]

How much would you say is optimal?

 

'Optimal' will depend on a variety of variables that aren't (currently) stated but can be anticipated.

This is somewhat equivalent to 'Style' in that abstraction of form generally serve specific functions.

Barring the understanding of what style will be targeted I'd say 'optimal' in this case is equivalent to 'something that can be easily animated'.

More abstracted targets need less optimization.

Why? Because if higher fidelity isn't required we are granted an allowance for greater tolerance.

 

But 'something that can easily be animated' is in itself a bit too abstract because what does 'easily animated' mean?

With CG characters this generally relates to rigging and deformation and 'optimal' is driven by what the character/object will be animated doing (form vs function).

Optimal for a prop that doesn't require any articulation has less need for optimization (but note... that may make it an even better candidate for simplification which can result in optimization).

In this case however we have a CG semi-realistic character that is expected to mirror qualities seen in real humans.

As such there is a higher bar to hit for quality and optimization.

 

But higher quality does not necessarily equate to more complexity.

And this is the aspect of 'optimization' I refer to above.

If the mesh is too dense (i.e. denser than required for the given task) then it works against the animator.

 

Optimization often leads to building multiple versions of the same model, some with more complexity than others.

Full optimization then would be to select the appropriate version for the right task.

 

An example: The character seen at a great distance might require 20 percent of the median mesh's geometry while that same character on a close up of the face might require additional detail be added over and above what is considered to be 'the the final model'. In other words, we may not know what is required until we note its absence.

 

But that's the long answer.

Here is the short answer : To obtain a more optimized semi-realistic model consider removing nipples from the female models (assuming they don't serve any purpose in that particular model). Consider removing other excess detail from other models as a means to further optimize. Note: If nipples are required to tell the story... or any anticipated story with that character... removing them will not be optimal.

[itsjustme]

 

lower the density of the mesh created with MakeHuman from 'ridiculous' to something approaching 'optimal'

How much would you say is optimal?

 

 

The model Malo posted has a little over 20k patches. To me, manageable would be under 10k. An optimal number would depend on the model.

[Fuchur]

10k-15k at most I would say. (15k will very likely result in longer Refinding Normals times)

 

See you

*Fuchur*