Newsgroups: comp.speech
Path: pavo.csi.cam.ac.uk!doc.ic.ac.uk!warwick!pipex!howland.reston.ans.net!agate!library.ucla.edu!news.mic.ucla.edu!unixg.ubc.ca!acs.ucalgary.ca!cpsc.ucalgary.ca!hill
From: hill@cpsc.ucalgary.ca (David Hill)
Subject: Re: Synthesised singing
Message-ID: <CJ4nq6.Arp@cpsc.ucalgary.ca>
Sender: news@cpsc.ucalgary.ca (News Manager)
Organization: University of Calgary Computer Science
References: <9400411.7766@mulga.cs.mu.OZ.AU>
Date: Tue, 4 Jan 1994 22:56:29 GMT
Lines: 62

In article <9400411.7766@mulga.cs.mu.OZ.AU> bromage@mundil.cs.mu.OZ.AU (Andrew James BROMAGE) writes:
>I'd like to know if anyone has ever thought of synthesised a human voice
>singing? Has anyone tried it out and how did it turn out?
>
>I thought of it, but I can't afford to record my own phonemes and the
>ones available have pitch inflections. (I don't know if that's the
>correct term, but you understand what I mean.)
>
>Post or email, I don't care.
>
>Thanks,
>Andrew Bromage
>-----------------------------------------------------------------
>Andrew Bromage			| "When the going gets weird, the 
>				|  weird turn pro." 
>bromage@mundil.cs.mu.oz.au	|	- Dr Hunter S Thompson 
>bromage@ecr.mu.oz.au		|   X <- You are here 
>If any opinions expressed here match those of the University of
>Melbourne, I'll sue them for plagiarism.

You could try Perry Cook at CCRMA

CCRMA, Dept. of Music
Stanford University,
Stanford, CA 94305
Telephone: (415)494-1533
<name>@ccrma.stanford.edu

OR Xavier Rodet who was at IRCAM last I knew
IRCAM
31 rue St. Merri,
75004 Paris
France
email: <name>@ircam.fr

You need something better than a consumer synthesiser chip to do singing.
Rodet did time domain synthesis, using the sum of a series of formant wave
synthesisers, and controlling the attack and decay of each.  The quality
of the singing was incredible (try his Aria from the Queen of the Night,
by Mozart).  His system was called "Chant".

Perry Cook has been working with waveguide synthesis (tube models) and
has also done excellent work.

I have played around with a formant synthesiser (PAT, which was very similar
to Klatt's formant synthesiser, and can be simulated in real time on a DSP
chipd these days).  I found synthesised singing was very much better than
synthesised normal voice, mainly because the rhythm and intonation are
well specified in singing (although much work remains to be done to see
exactly how singers value the notes, and exactly when they change pitch, how
fast, and so on).  But when there is a melody as well as speech, there is
often more redundancy, because people may recognise the music and know
the words, or at least the subject, so their ability to understand the speech
is enhanced.

Anyway, that's a start.  Please keep in touch.

david
-- 
david hill: hill@cpsc.ucalgary.ca	|	Imagination is more
voice: 403-282-6481, fax: 403-282-6778	|	important than knowledge.
nextmail: hill@trillium.ab.ca		|		(Albert Einstein)
