1/21/2024 0 Comments Jolene 33 rpm pitch shift![]() try saying "boot" when your mouth is configured for the "eeeee" sound!!! but rather you configure your vocal system to make the "eeee" sound and then overlay "b" and "t". So, "beet" is NOT a "be", then "eeeee", the "tuh". We defined a spoken word as an underlying vowel sound with the other elemnts (e.g. And then more, of course, as this had enough info for the Prolog templates which defined the voice. The next stage was to do voice modelling, which we could do given a sample of the person's speech - didn't take it too far as this was not the prime aim of the research - so, it was confined to "vowel-y" sounds - so we could accurately get the system to say "I owe you a yoyo Roy", given that sentence from the person. In the end we could produce speech which on statistical testing showed that people could not tell it from the real thing - unlike many pure synthesis systems (ie NO samples at all). ![]() The front end was written in Prolog by the Linguistics/Phonology folk, and I did a graphical interface into the generated "code", in order to "fix" and "understand" things when they were not "quite right". Many years ago I was working on speech synthesis (YorkTalk), which used the same engine as MITtalk, Klattalk, DECTalk (and variants), originally dveloped by Dennis Klatt. Isn't there software yet where you can take a sample of someone speaking and then type text in and have "them" "speak"?ĭoes only the CIA have this right now or something?
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |