Sorry, I made a mistake. Some time ago when I was looking at festival,
I noticed a webpage describing the use of festival to synthesise sung
voice. Upon seeing your post I did a quick search and posted what I
thought was the same thing. My description of what to look at in the
quoted page was wrong, as the festival bit down the bottom (even with
their new voices) is normal text to speech.
After noting I'd sent you off-track, I had another look. Flinger is the
system I was thinking off:
http://speech.bme.ogi.edu/tts/flinger/
But again I haven't used it or looked deeply into it myself. So I'll
shut up now before I dig myself deeper into a hole :-)
Cheers,
Ross-c