by MTKnife » Thu Mar 18, 2021 4:18 am
I can be a little more specific. There are two issues with creating characters under 25, one of which you can work around rather easily, and the other of which you would have more trouble dealing with. As Rob hinted at, both of these problems result from the fact that the program uses linear interpolation between different target models. To put it in greater detail, the program doesn't use some sort of formula to represent a human of a given age, sex etc. Instead, there are 144 (if I'm doing the math correctly) "target" models for each combination of gender (male/female), age ("baby"/"child"/"young"/"old"), muscle, weight (with a min, max, and average for each of these two), and height (min and max). If you pick a sex, age, muscle mass, weight, and height--using the sliders provided in the GUI--the program interpolates between two or more of these targets.
Most of that works relatively well, but age, unfortunately for you, is the worst handled of the 5 sliders. For these targets, the "baby" model is a 1-yr-old, the "child" model is 10, the "young" model is 25, and the "old" model is 90. The choice of these four numbers (especially the 10 and 25), combined with the interpolation logic, causes the two issues I mentioned above.
The first problem results quite straightforwardly from the interpolation logic: if you create a character between 10 and 25, the character's height will be a distance between the heights of those two ages that's proportional to the distance between those two ages. Thus, if an average female 10-year-old is 48 inches (I don't remember the exact numbers, but you get the idea), and a 25-year old is 66 inches, a 15-year-old is 54". In other words, height changes at the same rate as body shape. That's obviously wrong, since women tend to reach their mature height around age 14-15--and even before that, most of the growth is from 10-13 (according to a source I just looked up), while their secondary sexual characteristics mature more slowly (that is, they keep getting bigger breasts and wider hips long after reaching maximum height). What's more, the program uses the same formula for boys, interpolating between 10 and 25, even though boys mature a couple of years later. This particular problem is the one that drives me personally crazy, but it's easy enough to work around, by adjusting the height slider to produce a realistic number.
The other problem--and it's the one that will most concern you in particular--is more complicated. I don't spend a lot of time examining prepubescent bodies, but my understanding is that male and female body shapes are similar before puberty begins, with the obvious exception of the genitalia. Even if the targets are correct (and they look OK to my untrained eye), by 10, puberty has already started for most children, especially girls, and so the child targets show some body shape differences owing to differences in secondary sexual characteristics. That's fine if you create a kid older than 10--other than the height, the interpolation between the 10-year-old and 25-year-old targets produces a reasonable result, assuming of course that the body changes in a more or less linear fashion between these two ages. Ergo, as long as you don't need to create any models under 10, you should be able to get at least close to what you want, though I would imagine that most people's bodies change faster between 10 and 18 than between 18 and 25. (And of course, keep in mind we're talking about averages--the 10-year-old target may be appropriate for some early-developing 7-year-olds and late-developing 13-year-olds.) However, if you do create a child under 10 (which, to be honest, probably isn't done that often with Makehuman), you run into a problem: the interpolation means that differences in male and female body shape are going to be noticeable even in very young children--and the distortions will be greater for girls, because the 10-year-old female targets are more mature than the 10-year-old male targets. As Rob said, you could certainly fix this by adjusting a number of modeling sliders by hand, but it would be a lot of work.
When the program was first coded, different "child" ages should probably have been picked for males and females, maybe something like 8 for girls and 10 for boys. And I really don't know why there are even different male and female baby targets, since the program doesn't even include genitalia (though that can be added with downloadable content). At this point, though, someone would have to redo 18 female child models to fix the problem, and that's not in the cards (though I wonder if you could put the girls' heads on the boy child bodies, and add slightly different interpolation logic?).