well thats a good question… i have wondered about that, too.
i’d say your best bet would be to duplicate the model and slightly scale it inwards along the normals, then adding hair particles to that instead of the model you’re actually going to render und applying an invisible material to the duplicate model.
i don’t think having a translucent skin is a very good idea.
but SSS might allow you to see some of the origin through the skin because it should create a sort of ‘internal shadow’
i have actually done this for my old geezer, but i suppose the SSS is to weak and the duplicated model isn’t scaled in enough for it to be noticeable.
i’m trying to be subtle with SSS but now i’m thinking it might be better to increases it quite a bit, but add an actual modeled skull inside the head to prevent it from affecting the parts of the head that should be bony.
i’ll do a quick test once i get back home from work and post a close up shot.
about the filesize… well it depends on the amount of quads you’re using obviously… on the higher end it can reach about 50-150mb for a single model.
i think the old geezer uses up ~40mb atm, but i’m getting all the detail i need. just one subdiv higher and it would probably be at ~160mb tho.
using a low poly + subdivision + displacement map you can shrink it down to something smaller than 5mb