That sounds like a very good idea.. or I could force the models to all have an _lm suffix in MEL with a renaming operation.. though then there would be duplicate models-- using the _lm suffix in the splitter is probably a more efficient idea.

Actually the really hairy part is generating the c-script to load all the different _lm meshes and their bmaps.. can you think of a good strategy to automate that? With so many objects, that is the restrictive barrier.

Thanks for the advice!