(Exerpt from an email from Paul B.)
This place is looking to create a free text-to-speech:

http://tcts.fpms.ac.be/synthesis/

http://tcts.fpms.ac.be/synthesis/mbrola/mbrtts.html

We can probably get a way to integrate that into Java.  Here's some potential code:

http://sourceforge.net/projects/comedia http://freetts.sourceforge.net/

I don't particularly like the voices that they've developed and it looks like you have to use a very odd phoneme format of text to get the sound which may or may not be possible to automate.

When I did the demo for UDL I found some good standard MacOS X voices for Spanish from Cepstral.

http://www.cepstral.com/cgi-bin/demos/general

That company has voices available for English, French, Italian, German, and Spanish.

Basically, we would have to buy the $300 SDK and figure out how to integrate it with Java. Then to actually release things using those voices it would cost us a couple thousand dollars for 10 languages.

Here is another similar company:

http://www.acapela-group.com/demos/samplesHD.asp

So, doing text-to-speech is going to cost a lot of programming time and possible licensing fees. I'll check with Scott a bit about what he knows about Java TTS, but we might be better off for these example lessons to just use spoken phrases.  Real spoken phrases are going to give better quality than any TTS at the cost of making it very difficult to subsequently change the lesson without re-recording everything.  If this part were dropped in as one of the last steps, recorded phrases would work well.