I have low-poly character models that make heavy use of textures to define shape using the alpha channel. (An example would be glasses, which are only 3 solid large quads, but the alpha channel of the texture defines the shape of the glasses)
So there's a tonne of transparency/translucency goin' on, I estimate that one quarter of every tri drawn, will have at least a little transparency in it's texture needing blending.
Most of the transparency occurs in the hair, which layers up geometry very close and intersects itself sometimes.
My question, is there any way to get reasonable alpha blending behavior, without ordering which tris are drawn according to depth? Most (old) opengl documentation I've read says ordering is required to blend any kind of transparency, but I feel ordering every tri every frame, in this case, would be too costly. (or at least, beyond my ability to code without spending weeks on it)
There is no lighting or vertex coloring, if it's important. The results don't have to be accurate, but I would like them to be visually pleasing and consistant, if possible.