I first heard about a project, at Microsoft, I think, about twenty years ago that used recordings of a specific person's voice as the basis for a synthetic voice. In that work, the person had to record a set of utterances that covered all necessary phonemes. I believe they also got reasonable results from recordings of historic figures, but it's more work. At the time, I speculated that if we ever had AIs on space ships they would sound like Majel Barrett because all of the Star Trek fans would expect that! I'm glad to see that they are actually trying to do this now. I think Mrs. Roddenberry would be amused.
1 comment
[ 3.1 ms ] story [ 9.1 ms ] thread