The voices are all just autogenerated from text, including the one we distribute. We just only generate one, with a particular set of settings for a particular text to speech engine. If you generate your own (which the rockbox utility can do) then you can use any text to speech engine we support that you have installed, using any settings you like.
There isn't a repository of voices that I know of, because the voice file needs to be updated whenever Rockbox changes the text strings that it uses, and so any pregenerated voice files would go out of date very fast and when that happens they start saying the wrong things.