Greetings, I posted this over on reddit blender, and thought I would share it with the makehuman community to.
I started out playing around with makehuman, its actually rather flexible with what you can create, I believe makehuman is a great time saver as it saves you potentially years of creation time, and gives you a great starting point. But that's not to say you shouldn't learn the basics of character rigging first.
I like how the developers thought of allowing custom assets, as makehuman out of the box is rather simple, for example, you cannot create a realistic nose or cheeks, so you need to create many different custom targets that can be used to better manipulate the makehuman mesh, more so than the default sliders.
The default UV layout is rather nice and easy to work with, however the head/face UV area is terrible, this is because the pixel density around the face is only 260x260 pixels, in my opinion this is not enough pixel density to have unique facial features once rendered with SSS as even an area of 512x512 would still be to small now days. Using the default skin I was never able to get away from the classic makehuman appearance. So I decided to change the head island, making the back of the head UV area much smaller, seeing that its most likely going to be covered with particle or mesh hair anyway that detail is not important and may as well be sacrificed for greater face detail. I also decided to increase the overall skin texture resolution to 4096x4096 giving me much greater overall resolution, with a face texture area of around 880x880 pixels.
This brings me to the most frustrating part about this characters development, the Transfer UV feature does not reassign or rebuild the vertex order, (maybe there is a good reason for this). But after an entire day of reading over forum posts about the same issue, I came across this little gem about Sort Elements, I can not express how important this little mostly unknown blender feature is, specially when you have a number of different master models, setup with different UV layouts, for projection painting, that take days to setup. (Actually, sadly, sort elements turned out to be completely useless for matching vertex orders, so I have been learning how to create custom proxys, so that I can more easily use my new UV layout).
Shaders where learned over about 2 years with lots of research, I learned 100's of different materials over this time, not just human shaders. At first I did a few tutorials but soon realized I would need to go much deeper into learning shader nodes. Some tutorials are a decent starting point, but then a lot of tutorials will teach you something that will only look good in a specific lighting environment, making them completely useless when used on a character or in a proper lighting environment setup.
The skin shader uses a diffuse, light map and normal map. The skin texture has enough resolution to retain skin pores and other facial imperfections, but the skin pores don't really show up unless the camera is extremely close to the face, and I render at around 2000 samples. For this character I decided to leave out a lot of imperfections and just go with an airbrushed look. For my next character I may step up the face texture resolution to 2048x2048 and give the face area its own texture altogether.
The hair and eyebrows are particles, setup in a way that best recreates how real life hair grows using around 6 different vertex groups. The eyebrows were very difficult to make, and are in their 4th generation, im still not %100 happy with how they look, but I think I have the shader mostly worked out, as they look soft and natural.
The lips are setup using a mask, so that they can have a different glossy appearance to the main facial area, however im still not really happy with the transition between the skin area and the lips, any advice on how to improve this would be much appreciated. There is also a small mask around the inner eyelids done much the same way, to try and give the eye lids subtle wet look.
Total creation time would be around 2 months on and off in my spear time, spanning out over around one year, including all the first attempts through trail and error, and also all of the improved shader experimentation. I know what people mean when they say they're never really happy with the end result, and putting so much time and effort into something can leave one feeling empty.
Shout out the blender developers, makehuman developers and HDRI Haven.
I welcome any tips on ways I could improve, I'll also try and answer any questions people may have.