odf opened this issue on Oct 27, 2008 · 13933 posts
MikeJ posted Thu, 22 October 2009 at 12:24 PM
Quote -
but then, i also hate how the difference in body and face scale means that a really detailed face will show up right next to a completely unacceptable in raw renders neck and shoulder for most portraits.
Just one of those things.
Since the whole idea behind good UV mapping is to try to eliminate distortion, programs that do it the way UVLayout do it create UVs that are proportional to the polygons they correspond to. In other words, if you don't alter the size any, a properly unwrapped head polygon will be the correct size, proportionallly, to a foot polygon.
In the case of human figures that means that if mapped together and left proportional, the head will take up a very small portion of the 0-1 UV space, maybe only 1/7th or 1/8th of the size of it. So, even using a 4K texture map, that limits you to 500-600 or so pixels for the head, and that's as good as it gets. Use an 8K texture map, and it's somewhat better, but then you have 40 MB texture maps using half a gig of RAM to render.
So then also, your render shouldn't be larger than the textures that were used, or you get pixellation. In other words, with a head proportionally mapped to a body, even with an 8K texture map, you wouldn't be able to render a face shot any larger than about 1-1.5K.
By making the head (and other body parts) take up as much of the UV space as possible, you throw that proportion out of whack, but you also make much larger renders possible without risking pixellation.
So it's a trade-off. And that difference between scale doesn't have to be noticeable either, but it takes some work. Same goes for the lips.
I've already mentioned several times in this thread that the better way to UV map is to use multiple regions, where every important part can take up a whole UV square, or at least be very large, and Poser 5+ can handle it too, since the material room allows for setting tiling UV regions.
I would personally prefer it that way, but then you start getting numbers like 0-1, 0-minus1, 1-2, and so on, depending on which direction your regions go.
The problem, of course, with that is it could create problems for people who don't know how to deal with textures in that way.
So I did it in the traditional DAZ sort of way - overlapping and resizing.
But this is why I keep asking for feedback and opinions - to see if the texture people in this project want something different. Many things have already been changed, and I'm willing to make as many changes as needed until it's "just right".
Unfortunately though, there isn't a way to have it one way and many ways at the same time. Not until they invent QuantumPoser Pro 2050. ;-)