Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!rutgers!ucla-cs!sdcrdcf!trwrb!cadovax!keithd
From: keithd@cadovax.UUCP
Newsgroups: comp.sys.amiga
Subject: Re: Questions
Message-ID: <1386@cadovax.UUCP>
Date: Fri, 6-Feb-87 15:29:01 EST
Article-I.D.: cadovax.1386
Posted: Fri Feb  6 15:29:01 1987
Date-Received: Sun, 8-Feb-87 05:46:14 EST
References: <775@vaxb.calgary.UUCP> <34@f.gp.cs.cmu.edu>
Reply-To: keithd@cadovax.UUCP (Keith Doyle)
Organization: Contel Business Systems, Torrance, CA
Lines: 43
Keywords: Visual 550 emulator wanted, Caligari status?, memory expansion

In article <34@f.gp.cs.cmu.edu> mjw@f.gp.cs.cmu.edu (Michael Witbrock) writes:
> I would like to suggest that the next release of intuition support a
>narrator device with a full set of IPA phones. This will make the usefulness
>of the machine in educational markets greater as it will be able to
>pronounce foreign languages. It will also mean that those of us who find
>other accents euphoneous will be able to have our machines speak them.

I had this interesting idea, and in fact now have enough software tools
put together to at least test the feasibility of producing a narrator
device that does it's speech by playing back sampled phonemes instead
of synthesized ones.  

Though I can now cut and paste sampled 'phonemes' fairly well, and even
'drive' them from a text string, I have yet to actually attempt to accumulate
the required quantities of phonemes in order to actually hear how it will
sound.  One problem, is I don't even have the best model it would appear
from your posting, how can I find out what the full set of IPA phones
are?

And there are other potential problems.  I'm not sure really how to
'meld' from one phoneme to the next effectively if it turns out that
simple splicing is not good enough.  And even if simple splicing IS
good enough, many of the phonemes are so short that cutting them
out by listening can be pretty tough.  A phoneme taken out of context
begins to sound like nothing, or noise, or whatever, especially when
you're listening to it over and over while you cut and paste. 

I've been wanting to take the soundtrack from one of these UFO movies
I have on video which was narrated by Rod Serling (and maybe a few old
TZ's) and see if I could accumulate enough Serling phonemes to come up
with a speech program.  I realize there are all sorts of copyright problems
with such a program, but I want to first find out if it is even feasible.

Unfortunately, there's a lot I don't know about speech synthesis and
phonemes etc., and it's been a back-burner project anyway, but since the
subject came up, I'd like to know if anyone out there knows any good
references on the subject that might help out.

Thanks,

Keith Doyle
#  {ucbvax,ihnp4,decvax}!trwrb!cadovax!keithd
#  cadovax!keithd@ucla-locus.arpa