I don't know how promising voice reproduction is, but I do know that there has been a shift in the basic theory behind it in the past 20 years. I previously worked with a group that was given the job of writing the software for Steven Hawking's voice. It turns out that Steven Hawking's voice reproduction is done by an ISA voice card from the early 80's. He now is provided with regular upgrades to the computers and hardware on his wheelchair. The only problem is that these cards cannot be found anymore and he only has 3 cards. So if it is no longer possible to support these cards or if they break, then he will have to use a new voice synthesizer. But Dr. Hawking does not want to have a different voice (the cards do have a subtle uniqueness to them) and so they were trying to get companies to write software for his voice reproduction that sounded the same as the cards. Apparently this was hard to do because the basis of voice reproduction has changed since the early 80's and nobody really knew how to reproduce the same effects with the new algorithms. I do not know how to explain the difference though but it makes for a mediocre anectdote.
