home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.ibm.pc.soundcard
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!magnus.acs.ohio-state.edu!cwatters
- From: cwatters@magnus.acs.ohio-state.edu (Coyt D Watters)
- Subject: Re: Sbos!=SBPRO
- Message-ID: <1992Dec17.195655.23774@magnus.acs.ohio-state.edu>
- Sender: news@magnus.acs.ohio-state.edu
- Nntp-Posting-Host: top.magnus.acs.ohio-state.edu
- Organization: The Ohio State University
- References: <1992Dec16.164820.538@mfltd.co.uk> <1992Dec16.185301.7654@magnus.acs.ohio-state.edu> <bjornhk.122.724616844@dhhalden.no>
- Date: Thu, 17 Dec 1992 19:56:55 GMT
- Lines: 38
-
- In article <bjornhk.122.724616844@dhhalden.no> bjornhk@dhhalden.no (BJORN HAAVARD KLEVEN) writes:
- [me writes]
- >>Since this is a wavetable card, how hard would it be to make samples of
- >>all the allophones (phenomes, whatever) which make up language and put them
- >>together to do synthetic speech?
- >>
- [snip]
- >>I want more flexibility than simple playback of digitized speech, but I
- >>want to approach the quality of it.
- >
- >The problem which speech-synthesising, see, is the text-to-sound conversion.
- >Sampling the phonemes simply isn't enough. Just think about the words
- >
- >thanks. thanks? thanks! thanks... thanks!?!
- >
- >By using the 4 phonemes this word needs (I won't attempt to transcribe them
- >here..:-), all the five words would sound exactly the same, although they
- >have different stress. But that's only a small problem - the punctuation
- >tells us what we need to know. The tough(er) part comes when it comes to
- >different pronounciation of the same letter, or of the same sets of letter,
- >even. Consider: 'laugh' and 'ghost' the 'gh' part is spelled the same way,
- >but it doesn't sound at all the same when it's spoken. Now, these problems
- >were just a couple that came to mind (the SB talker included with the SB has
- >actually solved both), there are a lot more, and not an expert. It is
- >possible, but it'll take a lot of work.
-
- Hmmm. Must be the text-speech software. My Currah's software handles
- Ghost and Laugh pretty well. It also featured a "dictionary" which allowed
- for customized pronounciation. Macintalk was a LOT worse than the
- Currah Cart on my C64. The Currah's software also did pretty good with
- punctuation, although averything else would be spoken in a monotone.
-
- It did, however pronounce ghoti as G o (long o) tee.
-
- Of course, if SBtalker solves it.....(big grin). Will SBtalker work with
- SBOS?
-
- -Coyt
-