home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!andrew.cmu.edu!<UNAUTHENTICATED>+
- From: mjw+@cs.cmu.edu (Michael Witbrock)
- Newsgroups: comp.sys.amiga.multimedia
- Subject: C= : Where's the Speech?
- Message-ID: <If_atQu00hsB4p7WdZ@cs.cmu.edu>
- Date: 12 Dec 92 12:57:32 GMT
- Article-I.D.: cs.If_atQu00hsB4p7WdZ
- References: <Bz1v4z.4qI@undergrad.math.waterloo.edu> <Bz276t.n1x@news.iastate.edu> <57432@dime.cs.umass.edu>
- <Bz48IL.2su@news.iastate.edu>
- Organization: Carnegie Mellon, Pittsburgh, PA
- Lines: 33
- In-Reply-To: <Bz48IL.2su@news.iastate.edu>
-
- Excerpts from netnews.comp.sys.amiga.multimedia: 11-Dec-92 Re: C= :
- Where's the Speech? Dorc N. Barrenn@iastate. (4172) [or someone with a
- name like that]
-
- > In other words, the speech is good enough already that I just don't see all
- > that much room for improvement. You can install an array of DSPs or even an
- > array of Crays all dedicated to doing speech synthesis, and the speech would
- > not be improved all that much.
-
- The prosodic model sucks. Coarticulation isn't modelled well, rate
- dependent phone deletion doesn't happen properly. The synthesiser can't
- speak any language but American English. The source model is probably
- white noise instead of modelled glottal pulses. The text to speech
- algorithm makes frequent errors. I defy anyone to listen to the Amiga
- read their mail to them without needing to look at it as well.
-
- The speech is far from perfect. There are better synthesisers out there,
- and plenty of room for improvement. Anyone who has heard a human being
- speak should know that the speech can be improved a lot.
-
- That not withstanding, amiga's speech synthesis is pretty good as
- non-research systems go, and has such an important place in Amiga
- software that it needs to be kept and improved. If a serious effort were
- made to support multiple languages, amiga speech would be a uniquely
- useful product.
-
- I'm not sure that speeding up the model with a DSP is the only way of
- improving the speech, however, - switching to something like
- interpolated demisyllable synthesis instead of the current source/filter
- articulatory model might help more.
- ---
- In any case, rest assured that Marc knows nothing about speech synth,
- but that the Amiga needs it.
-