I've uploaded a new test version of eSpeak at http://espeak.sourceforge.net test-1.09e-linux.zip (27.April.2006) I've included improvements to the speaking of number strings, following the comments here, but there may still be some peculiarities. Let me know :-) I've now changed over to using UTF-8 encoding rather than Latin-1 for eSpeak's data files. It now accepts (and expects, unless it notices otherwise) UTF-8 encoding for its text input. This should make no difference for US ASCII text, of which UTF-8 is a superset. For text with accented characters or symbols such as Pound or Copyright, eSpeak should be able to recognize whether the text is UTF-8 or Latin-1.
