Path: utzoo!utgpu!cunews!bnrgate!brtph3!brchh104!brchs1!bnr.ca!rice.edu!sun-spots-request
From: cfreese@super.ORG (Craig F. Reese)
Newsgroups: comp.sys.sun
Subject: Re: Making your machine talk...
Keywords: Miscellaneous
Message-ID: <685@brchh104.bnr.ca>
Date: 8 Dec 90 22:07:00 GMT
Sender: news@brchh104.bnr.ca
Organization: Sun-Spots
Lines: 23
Approved: Sun-Spots@rice.edu
X-Original-Date: 5 Dec 90 21:00:07 GMT
X-Refs:  Original: v9n364, Replies: v9n381
X-Sun-Spots-Digest: Volume 9, Issue 391, message 15
X-Note: Submissions: sun-spots@rice.edu, Admin: sun-spots-request@rice.edu

In article <513@brchh104.bnr.ca> jsb@cs.brown.edu (John Bazik) writes:
>In article <292@brchh104.bnr.ca>, jack@cwi.nl (Jack Jansen) writes:
>|> I am trying to build a program that talks to me.

>The obvious hack is to record all the phonemes and write a backend to
>string them together and write them to /dev/audio.

I tried this with very limited success (I didn't expect it to work well
anyway).  Here's what I did:

   - Use NRL text->phoneme algorithm
   - segment the phoneme stream into individual phonemes
   - paste together speech segments from digitized recording
     of the phoneme "words" (i.e. beet bit gate...)
   - ship the result to /dev/audio.

It is a fun excercise but don't expect very much.  Its not as good as my
$60 text to speech board that I plug into the RS-232 connector.  If I can
find the time I think it could be made better with some signal processing
hacks.  If anybody has a better version I'd very much like to hear (no pun
intended) about it.

Craig F. Reese                           Email: cfreese@super.org