Hi all,

This is just something I'm curious about actually. Do people here prefer to go on ahead and model/sculpt a human character from scratch or use some sort of baseline mesh generator program (like MakeHuman) and tweak it a bit? Why one or the other? I'm especially curious to hear from people who are indie and from people who are in the industry.

I prefer to use the mesh generators mainly because I'm a one man team, and doing this makes my life significantly easier, makes my main modeling go much faster, and lets me put finalized pieces out much quicker onto my website.

As a solo dev I always try to take advantage of exiting assets when possible to save time. I'm trying to build entire games, not just models.

I don't think any professional artist would use MakeHuman or any similar software as more than a reference, maybe only if they are pressed for time.

Art is about showing your own interpretation of a object, using pre-made meshes tends to prevent the artist from reaching your own goals.

Using any pre-made assets is considered the same as using a other artist's work along with yours and then claiming that it's yours, professionals avoid this so that the work can remain there own.

One of the largest complaints people have with 3ds Max is that more than half it's download size is pre-made assets, a pointless waste as most max users never use them beyond testing and can make there own test assets; how many AAA games have you seen that uses pre-made Max assets?

The indie community I believe uses pre-made assets a lot more, because of the small teams that have to work on each game, however the better games amongst these games are the ones that adjust the assets to match the game.

As a solo dev I always try to take advantage of exiting assets when possible to save time. I'm trying to build entire games, not just models.

Yea I definitely feel you there. I actually really like modeling, but I want to have more than just models, even if I'm just going for a CG video, I want to get around to finishing it. Usually when I model, I don't mind spending time on assets that aren't human characters. For human character, I tend to use shortcuts (pre generated meshes, etc.) to cut down time, I end up modeling armor myself though. In the end, I prefer having modeled most of the stuff to keep my own visual aesthetic for everything.

I don't think any professional artist would use MakeHuman or any similar software as more than a reference, maybe only if they are pressed for time.

Art is about showing your own interpretation of a object, using pre-made meshes tends to prevent the artist from reaching your own goals.
Using any pre-made assets is considered the same as using a other artist's work along with yours and then claiming that it's yours, professionals avoid this so that the work can remain there own.

One of the largest complaints people have with 3ds Max is that more than half it's download size is pre-made assets, a pointless waste as most max users never use them beyond testing and can make there own test assets; how many AAA games have you seen that uses pre-made Max assets?


The indie community I believe uses pre-made assets a lot more, because of the small teams that have to work on each game, however the better games amongst these games are the ones that adjust the assets to match the game.

I understand the notion of art being an interpretation. It's the reason why I generally model all of my assets except for the actual human mesh itself, because of the amount of time it takes to make the mesh.

So do people not even start from a baseline mesh that you may have modeled earlier? I've done character modeling for a class, and it just takes a while imo. How long does character modeling take for a professional? On average, that is.

So do people not even start from a baseline mesh that you may have modeled earlier? I've done character modeling for a class, and it just takes a while imo. How long does character modeling take for a professional? On average, that is.

Base meshes tend to be made per project.

Depending on the engine, art style and what the model will have to do, a base mesh is formed to match all these requirements, often a male and female base is made.

As a example if you are using Unity and want to take advantage of it's inbuilt animation re-targeting, all characters will have to be in a T-pose; things like blinking eyes will also effect what the base mesh looks like.

Only basic NPCs are made from the base, special characters, monsters and animals all have unique proportions so the base isn't that useful for them. Think about the Batman games, you wouldn't make the basic thug, Batman and Penguin from the same base.

Most artist have a library of body parts that they use as reference, all these parts are from models they made in the past; the reason you don't use them to stitch a model is because even the rights to each body part transfers with the model.

Time for a AAA game character ranges from a week to three weeks per character, that is for the mesh, textures and rig to be made, animations take much longer as they are captured and processed; it's no different from movie scenes and take about the same time.

Very low poly character like the ones used in games take 3-5 days, and about 4-6 hours per animation to make; you will want to aim for something in between for a indie PC game.

It also takes more than a year of studding anatomy before you can make characters that look good.

This is something I'm a bit curious about as well. I'm currently enrolled in an art program, where I expect to be modeling human characters from scratch before it's all over. But I've looked at some of the programs like Make Human. And I've even heard people who are more in the know than me talk about using things like DAZ studio and such. I think there are advantages and disadvantages and I'm still trying to learn what they are.

For indie developers who are more game developers than artists, these seem like really useful tools to me. Even if you're an artist, what if you need 400 individual human models for the game!? That's a lot of work. Having a program that can crank them out quickly may make some sense. I mean, think of Grand Theft Auto. There's all sorts of human npc's that you come in contact with that are not main characters. You may want to spend some time on important characters, but in a case like this you need to crank them out in volume. Seems to me that a program that can help you with that would be useful.

I've only done 2 or 3 human models in my life so far and they were all following tutorials. Taking the artistic approach of actually modeling them, it would seem to me that you could get a base mesh fairly close that you could then use as a template so that you would not have to start over from scratch for each character.

I've also wondered about games where the player designs the character like Skyrim or the Sims. Seems you would have to have some artistic skill to create that. Make Human or DAZ studio or such are not going to help you there. There you need the artistic skill to kind of create your own "Make Human" inside your software.

I've been thinking you could standardize vertex positions where the body parts connect. For example, if you make it a rule to use the same ring of neck vertices, then you know that if you produce another head it will attach to the existing torso, allowing for character customization although not to the same degree as Make Human.

But if you just want a single character, like in Tomb Raider, then Make Human like programs make a lot more sense.

My experience with Make Human though, was not particularly great. Clothing and hair options are pretty limited. Plus, should clothing even be done this way? You're layering clothes on top of a body mesh, although there is some sort of option to hide the polys below the clothing. Does that actually delete the hidden geometry or just hide it from view? The models are probably not particularly optimized for games. I'm still a pretty beginner 3D artist, but I tend to want a very low poly count model with a good normal map produced from a high poly model. I doubt this would be much different. Make Human doesn't make that easy, if you can do it at all. I just used the overly high poly model. There's an option to export a low poly version, but I don't think it's conducive to baking normal maps and it looks pretty bad. Other programs may be better at this, but what I really need the program to do is output a low poly mesh that appears to be far more high poly than it actually is because of the normal map. Humanoids especially seem to me to be a good candidate for this technique. And clothing really looks far more realistic when you can add 3D wrinkles and such. You certainly don't want a high poly mesh in your game. Not for a commercial game anyway. Make Human appears to not even create a normal map for the body. (I see one for the hair.) I would prefer half as many polys and a good normal map that makes it look much higher poly than it is. I'd like 10,000 triangles with a normal map produced off a million triangle model.

I brought the high poly mesh into Blender and skinned it and rigged it by hand. This is good experience by the way. Make Human supposedly exports skinned and rigged. I tried exporting it every way I could to get it into Unity. The armature was all wrong. I had to export it with no armature, skinning, or rigging. Then create the armature and do the skinning and rigging myself in Blender. Then I was able to import it into Unity where it worked well enough for my purposes to learn game programming. That was the first time I did it. Seems like it might have gotten better at this. The last Make Human character I've got here looks like I may have got it to export skinned and rigged. It might still require going in and renaming all the bones in order to get it to import correctly into Unity's Mechnim though. I know a lot more about all this now, and maybe some of the troubles I had the first time around where just not knowing what I was doing.

This was awhile back, and the software may have improved since then. I was able to find my last Make Human model. She has 28,000 triangles with no hair. I would call that too many. I've got an older graphics card that is about to get replaced with a GTX1080. It can handle between 3 and 4 million triangles before it starts choking in Blender. Considering you have an entire game scene to render, you probably don't want to blow your entire poly budget on one character, even if it's the main character. I think it's fair to say this model is way too high poly for a serious game. To truly be useful, it should probably export a 10,000 or less triangle model with a normal map to make it look like a 1,000,000+ triangle model. But that's not what it does.

I thought there was an option to export a low poly model, but I just opened the program and don't see it. Seems as I recall the low poly model was too low poly. And that's not really what you want. You want a low poly model that looks high poly. That's done through normal maps, which it is not doing.

I'm looking at the topology. There's way too many quads in the toes and ears. But other than that, it's not that bad. I think you could flat out delete the back sides of the eyeballs, although not sure if doing that would mess up the UV map or not.

You could probably box select half of this mesh in Blender and delete it. Then mirror the half you deleted and redo this model in Blender with a pretty good starting base if you wanted to. With a lot of work you could probably go in and remove edge loops where they are not particularly needed. I might remove a loop or two from the foot, But I would probably remove all of them practically from the toes. The toe nails seem to be somewhat modeled as well, which could probably go. I might consider removing every other edge loop in the legs, arms, and torso. Or course, if you start doing this much work, you're going to have to completely regenerate UV maps and repaint the model. The more I look at this, the more I'm about half inclined to actually do it.

And to a certain extent, I think maybe that's kind of the idea behind Make Human: that you use it as a starting point for art projects. I think this model was designed for still picture art as much if not more so than games.

I downloaded DAZ Studio and paid for some base models just to create base meshes to use with 3D Coat.

I use DAZ Studio to tweak the parameters of the human mesh (like how muscular or fat the mesh should be), export to obj, import to 3D Coat as a voxel model and tweak more there.

Bam, a very usable base mesh to work with with a lot of details already on it (which differs to many other base meshes without those details).

I can now decide myself how much I want to rework those details or leave them unchanged, depending on what I do.

Sounds like the perfect compromise to me. I wouldn't want to use unaltered human meshes from makeHuman or DAZ Studio though, not for any important character anyway. I feel generally these are either poor quality meshes for various reasons, lack personality (especially the faces), or look way to generic (some of the meshes you export from DAZ Studio are amazing, but they tend to look very similar-ish).

This is were a quick art pass in a sculpting software of your choosing (to brush up the parts looking boring, like the face), as well as a good retopo (to make sure the topology is exactly what you need, and not built with movie making or posing stills in mind) can do wonders without taking too long.

Of course, given you want to invest time in an important character and we are not talking about fillers. But then same-ish looking filler chars are, IMO, a clear sign of a game where the devs didn't have enough resources for the scope they were aiming for. Might be just me, but filler stuff, no matter if characters or level props, that looks far less detailled and unique than the important stuff, just sticks out like a sore thumb to me.

That's the major reason that I like using these software: it's a good base starting point that I can then modify. Programs like ZBrush, 3DCoat etc. can more or less add the personality I'm looking for. Making a mesh from scratch would take me far too long to actually get stuff ready.

I'm honestly surprised that more people don't have some sort of base human mesh template that was sculpted to use for at least some human character models, with a bit of tweaking of course.

what if you need 400 individual human models for the game!? That's a lot of work. Having a program that can crank them out quickly may make some sense. I mean, think of Grand Theft Auto.

The way you achieve this is the same way DAZ did, you create modular characters and interchange parts. If your indie teamed scaled down to match the scale of the indie game, you could still achieve a lot.

It should be possible for a indie team with two character modelers to make as many characters as was used in the first "State of Decay" game, with in a year.

400 individual isn't realistic even for a AAA game, unless identical people with a different hairstyle is considered individual.

interesting side fact, most AAA games don't even reach 4000 individual meshes, it all gets recycled.

base mesh fairly close that you could then use as a template so that you would not have to start over from scratch for each character.

This is the point of the base, however because the base has such a large impact on the end product a easy way to get a unique character is to make a unique base mesh to start from.

It's also faster to make base meshes than it is to make a full character.

if you make it a rule to use the same ring of neck vertices

Only the amount of vertices used has to be the same, vertices are vector points so vector math can be used to stitch them to other vertices.

this is why the same head can be used on both a fat and a thin body.

Plus, should clothing even be done this way?

No for games it shouldn't. The reason why character costumes are named "Skins" is because it is part of the skin and to change it the whole skin mesh needs to be changed.

Sims uses stitching, so the head is a mesh and the body is a mesh and they are stitched into a new mesh when you exit the editor.

Also some options will divide the body mesh into more stitched parts like selecting individual pants.

There's an option to export a low poly version

The new MakeHuman has a lower topology option that is better than the older low poly mesh, a combo of math and a pre-made topology, however it's still not perfect for games.

The new MakeHuman combines there original workflow with work made by artist, so even the high poly mesh looks better now and renders faster. Years from now it could even rival what expert 3D modelers could make.

Might be just me, but filler stuff, no matter if characters or level props, that looks far less detailled and unique than the important stuff, just sticks out like a sore thumb to me.

If used correctly lower quality NPC characters can be a meaning full part of a game, in Fable 1 they made it clear where the player was and who could be killed without causing damage to game play.

Years ago I worked on a game with a student, where the player was a new arrival at school, all the other characters where very low poly faceless characters.

When you interacted with them, for example bumping into them, they would start increase in detail, like hair would be added or a face would be revealed.

The goal was to interact with a person until all of the characteristics where revealed. A mechanic like this can make good use of low quality models.

I'm honestly surprised that more people don't have some sort of base human mesh template that was sculpted to use for at least some human character models, with a bit of tweaking of course.

Like I mentioned changing the base is a easy way to change the character. Think of it as aiming a rifle, a small adjustment to the sights has a larger impact over a distance.

As for using software like MakeHuman and DAZ, it's the same as training wheels on a bike. At first it will assist you however there comes a point, when it only gets in the way of what you are doing.

@Scouting Ninja, Great post! Thanks! I'd be interested in seeing more on how that stitching works; I've always wondered how they were combining the meshes.