Is there any function or lib can do this?
If you are not talking about this: http://espeak.sourceforge.net/voices.html I can tell you:
It's a hard engineering work.. You have to study many subjects before doing or even understanding this. I don't think there's a library that does the thing by simply calling one or more functions, because unless the file is supposed to keep separate voice tracks, you have to work with waves, frequencies and complex integro-differential equations, and the result is not always automatically as good as you may hope..