i realise that this can be largely dependant on the engine you're using, and how it handles alpha channels and the like.
so here's the question: how do you handle hair?
how would you go about recreating FFXIII style hair? are you satisfied with the current "limitations" when making it? what do you think could be done better? do you have the solution?
this thread was made out of hours of frustration. but please, it's a serious post.
Replies
Edit* there was a link a while back on some behind the scenes stuff with FF XIII, i'll see if I can dig it up.
i find it gets more interesting with shorter/unkempt hairstyles. i once had to do sorta-unkempt medium-length hair (reference was tom cruise in vanilla sky). that was quite nasty to get to look halfway natural.
generally alphaplanes with tweaked normals (to blend to the underlying head-geometry) seem decent enough for the job. also keeping alpha to a minimum/only where necessary and trying to do most with solid geometry and texturework definitely helps. and it's important to not leave textureseams between planes and base. ever.
shading is another matter. be careful with too heavy normal map detail (also, seam-danger!) and i am not too sure on the benefit of hair shaders, particularly on longer hair. that sometimes looks way too much like out of some crazy unreal l'oreal ad. and again it tends to reveal seams.
http://wiki.polycount.net/Opacity_Map?highlight=%28hair%29
http://www.tylersart.com/thor.html
http://www.paultosca.com/varga_hair.html
I think showing a few ways depending on what you have to work with. So what hair would look like with basic alpha test vs alpha blend. We have plenty of peoples work to use as reference.
EDIT: I will say that the varga page dose the job of show the difference well, but having a few more examples of ways to archive nice hair with 8bit hair would be great.
In that thor tutorial the whole tile is Alpha mapped. would it not save a load of vram to split the hair onto a smaller individual tile or does the added draw call offset the gains
in other interesting news unreal is now able to shadow translucent surfaces. heres to praying bioware use this build in the new mass effect.
http://udn.epicgames.com/Three/ContentBlog.html#Dynamically%20Shadowed%20Translucency
In my experience, if the whole model uses the same alpha-blended shader then the fillrate is more expensive to the framerate than using two shaders... one w/out alpha blend for most of the model, and the other with alpha blend for just the hairy bits.
Using the same texture is good, removes another texture fetch since it's already there in video mem. Except as stated elsewhere that alpha is going to double the texture size (DXT5 vs. DTX1), so that might make it worthwhile to split into a separate smaller texture.
Two shaders does mean you're sending a separate chunk of vertices to the card, but it's less cost than all that alpha blend (usually, depends on how smart the renderer is).