From candide@urbanium.tv Fri Jan 04 07:57:38 2002 Return-Path: X-Sender: candide@urbanium.tv X-Apparently-To: lojban@yahoogroups.com Received: (EGP: mail-8_0_1_3); 4 Jan 2002 15:57:37 -0000 Received: (qmail 66902 invoked from network); 4 Jan 2002 15:57:37 -0000 Received: from unknown (216.115.97.171) by m4.grp.snv.yahoo.com with QMQP; 4 Jan 2002 15:57:37 -0000 Received: from unknown (HELO urbanium.urbanium.tv) (194.183.224.155) by mta3.grp.snv.yahoo.com with SMTP; 4 Jan 2002 15:57:37 -0000 Date: Fri, 4 Jan 2002 16:57:45 +0100 Subject: Re: [lojban] Re: Lojban Text to Speech Content-Type: text/plain; charset=US-ASCII; format=flowed Mime-Version: 1.0 (Apple Message framework v480) Cc: ioannis@urbanium.tv To: lojban@yahoogroups.com In-Reply-To: <20020104145445.GA1650@aerosol> Message-Id: Content-Transfer-Encoding: 7bit X-Mailer: Apple Mail (2.480) Received: from 212.68.238.171.brutele.be ([212.68.238.171]) by urbanium.urbanium.tv (JAMES SMTP Server 1.2.1rc2) with SMTP ID 108; Fri, 4 Jan 2002 16:57:35 +0100 From: Candide Kemmler X-Yahoo-Group-Post: member; u=92614944 X-Yahoo-Message-Num: 12770 > I suppose that works to just cobble something together. Still, > though, it will do nobody any good to be learning Lojban from a voice > of this quality; we'll all sound horrible the first time we actually > try to speak to another human. ;-) Which underscores the fact that > it would be really nice to have a high-quality, Lojban-specific set of > diphone recordings, which some kind folks in Brussels have recently > volunteered to do. If you can get your diphone list to them and give > them a hand in figuring out what to do, since you tried it once > already, it might make things easier for them. And you have confirmed > that you've already done all the work that I volunteered to do! Yep, no problem, I've checked: we have everything needed to record nice diphones: a good mic, a Mac G4 and an excellent sound program called SoundForge. So yes, please hand over the diphone list; we'll do our best. It'd be nice to know how to test the results as we go... > A few notes here. First, English vowels (esp. American English) are > very different from Lojban vowels, and for that matter from most other > languages. English vowels are not pure vowel sounds, but slide from > one vowel to another, as in a dipthong. Lojban vowels are pure, > crisp, and short. Hard to communicate the distinction in text, I > suppose. We'd send samples regularly, but again we have to understand how to make the TTS system use our recordings for that. > The Lojban 'r' is allowed to be pronounced in just about any way, > including the standard American untrilled 'r'. I happen to think that > Lojban sounds best with an Irish sort of 'r', but that could just be > because I love listening to Irish accents. ;-) I would advise > against the French "r", as that seems indistinguishable from "xr", at > least to my ear. Yes, I personnally cannot pronunciate "r"'s like the Italians do (trilled ?). And I don't know about the Irish "r". Perhaps the best consensus is the American "r"... Candide