there are already a fair number of really good computer generated voices out there. Problem is they cost like a zillion bucks to license and if I remember correctly, the hardware overhead is also fairly high.
So, as Susan mentions it's going to be a long while. The "sound ex" algorithm will be very complicated. Add to this the dictionaries need to be HUGE with a proper phonetic breakdown for each word. It can take years and years to build the lexicon/dictionary alone. But luckily both already exist.
I once wrote a neat "sound ex" for a LIMS I wrote (laboratory info management system) for an environmental lab. It was soooo "fun" dealing with all those chemical names. And this did not involve voices, just checking spelling of compound names. Still have the code somewhere on a floppy, I should try and find it to see if the disk is still good. BTW, it actually was a lot of fun and didn't take that long to complete, less than a month all totaled back in about 1993 using, of all languages, dBase III compatible code in the original Foxpro from Foxsoft.
|