You might see how easy it is to configure emacspeak the way you want
(it is readily configurable for types of voice speaking English; but I
don't know about the character-to-phoneme translation; that may be
buried and hard).
See:
http://www.cs.cornell.edu/Info/People/raman/emacspeak/emacspeak.html
You can pick up emacspeak via anonymous ftp from
ftp://ftp.cs.cornell.edu/pub/raman/emacspeak