Re: Lojban Text to Speech

["buzzwyrd" <jcrossco@bellsouth.net>]

--- In lojban@y..., Candide Kemmler <candide@u...> wrote:
> 
> So what do we do ?
> 
> - Do we let Ioannis (my Greek colleague) record the sounds with 
trilled 
> "apico-alveolar, fish-hook r's", knowing that some people might 
> experience problems pronunciating them ?
> 
> - Do we use the alveolar R ?
> 
> - Do we record both versions ?
> 

As long as your voice talent is willing, I say record them all and 
luxuriate in choosing.  The recording of diphones is the hardest 
essential part of the process.  Once you have good, self-consistent 
.wav files, you can play with them at your leisure. There will forever 
be ways to tweak and improve the voices, e.g., through clever 
intonation schemes and the like.  

And there's nothing wrong with having several Lojban voices, 
preferably of various dialects, genders, ages, pitches, etc.  Having a 
choice of voices would have been a significant boon when I was 
synthesizing the human-computer SHRDLU dialog.

By the way, once the voice(s) are available, someone could probably 
provide a web site to which Lojban text can be uploaded and 
automatically synthesized results then downloaded.

-Jack