Re: [lojban] Lojban Text to Speech

[Jay Kominek <jay.kominek@colorado.edu>]

On Fri, 4 Jan 2002, Bob LeChevalier (lojbab) wrote:

> >On Friday 04 January 2002 02:05, Jay Kominek wrote:
> > > The type of Linux and FreeBSD's /dev/audio isn't 'au', but instead
> > > 'ossdsp'. SunOS and NetBSD's type is 'sunau'. You can then leave off the
> > > -r, as it will be able to issue the IOCTL to the audio device indicating
> > > the appropriate sampling rate.
>
> In all the posts on this topic, other than the one wav file output, I have
> seen a lot of alphabet soup like the above.

The above was intended for someone who'd made the choice to use command
line tools.

> So where is the software you guys are talking about and if it isn't a
> download, about how much does it cost and where do you get it.

They're all using Festival, which is produced by the Centre for Speech
Technology research at the University of Edinburgh. It is only available
in source form from them. Linux binaries are available, but a cursory
search failed to reveal a Windows version.

> Nora uses Win98 on her computer.

The University of Colorado's Center for Spoken Language Research had, at
one point, a package for doing speech synthesis which simulated the human
vocal tract (and as such, you didn't record diphones, you described how
the vocal tract deformed for a given phoneme, etc). Their web page is
http://cslr.colorado.edu/ but I can't find that software anywhere.

If noone is able to locate it in a week or two, or can't get ahold of
someone via email, I'll go bug a relevent researcher in person. (I'd do it
more promptly, except I'm still on break and as such, avoid campus like
the plague. The research professors might not be back in their offices,
anyways.)

- Jay Kominek <jay.kominek@colorado.edu>
Plus ça change, plus c'est la même chose