I downloaded Read Please which works ok but it mispronounces nonstandard words badly and it still sounds like a robot. Are there any better ones?
I’m taking all of my classes online this year and it helps me retain material if I hear it aloud as I read it on the screen, so I’m experimenting with using a text-to-speech program to go through the online text “lectures”.
If you’re looking for something that doesn’t cost, you might try Festival, which is open source (see also Wikipedia). You might also check out the related Festvox speech synthesis, which aims to make it easier for people to implement custom voices. If you happen to use Linux, most distros I’ve used have precompiled packages available.
Alone, it will probably still sound like a robot to you; however, you can use the Edinburgh “speech tools” to add emotion expression and manipulation, which does a lot for making it sound more human.
We use Festival on some 1GHz laptops in our human/robot interaction research (I’m working on my PhD in computer science), as does almost every lab I’ve come into contact with. Just so you know it’s not crap software (although it may not have the bells and whistles you may find in a commercial package).
Not exactly what you’re asking for, but this is so impressive, I have to mention it. Vocaloid is a speech/voice synthesiser from Yamaha that can sing (and speak, apparently).