Thanks to the Opera browser, I recently (re)discovered the joys of Text-to-Speech technology. In particular, how good it is for proof reading. I was looking into VoiceML stuff and voice activated browsing on the Opera website, and downloaded the speech expansion pack. If you haven’t tried it out, you should because it is pretty rad. Anyway, I got it to read back the W3C Note I am working on and to my [not so] great surprise, it started skipping words, and sentences made no sense. Obviously, this was not a reflection on the quality of the text-to-speech engine, but to my own abilities as a writer. It seems that I suffer from some form of dyslexia.
I had also had some else read over the document, and they had not found as many errors as I found when proof reading with the text-to-speech engine. So it saved me a bit of embarrassment when submitting the document to be more formally reviewed by WAF-WG. I’ve now become a bit obsessed and decided to buy a more sophisticated text-to-speech engine that can integrate a bit more with Office and other stuff that I use. I opted for NaturalReader, which came with only one voice (Mike). NaturalReader uses the AT&T Truevoice technology, which I know nothing about. However, judging by the 500Mb voice file I had to download, I assume that it is not actually synthesising the voice, but probably has some sort of look up table for words or word parts.
Anyway, I’ve only been using it for about 4 hours and so far it has been pretty good. I don’t think it is as good as what Apple is about to come out with, but I believe it to be comparable. Maybe I will put up a sample, but in the mean you can listen to mike on the naturalReader sample page.
So! No more rereading over emails, word docs, and entries that I write 1000 times over. I just get the machine to read it back once or twice, make sure it makes sense, and send! I’m a happy chap.