BTW there is one cheat you could use for speech output: AT&T has a very professional Speech internet server. It has a html-form to enter the text, and then it will return the speech sound as a wav file to the browser. You _cound_ (although this may be not fully TOC conform) create the sounds on demand using a realtime TCP download from the server. But honestly, it may look strange if you app asks for web access, but has not internet functionality at all.