Composition-wise, I'd like the "singing" to be less Hatsune Miku, more Roomful of Teeth. The existing singing software I've seen doesn't seem like it would lend itself well to this, hence why I'm writing my own software.
Larger concepts I'm thinking about: Synthesized Gesture controlling Synthesized Sound. Musical Speech vs Speeched Music. Anthropomorphic Sound. Sound with a Face. Character Arcs as musical form through synthesized babel with expression?