Voxin was: Re: Switching to Linux

[Jason White]

Brandon McGinty-Carroll <bmmcginty at bmcginty.hopto.org> wrote:
> Janina and list, I would appreciate a list of any companies of names you
> know who are associated with this incarnation of IBMTTS.  It's worth a shot
> to see if we can contact one or all of these entities; doing nothing, as we
> (read that "me") have seen with speakup leads to a buggered situation.

I think this has been tried before, but if you know any good, well connected
lawyers who do pro bono work you could ask them.

First, there's IBM, naturally enough.

The original work was done by Susan Hertz at Cornell University, who I think
sold her company to Nuance. She has a Web site indicating that a new
synthesizer is under development.

Any of those parties could have obtained components from yet other companies
or individuals.

If you want to open up a formerly proprietary text to speech system, the ideal
candidate would be one which is not sold commercially, only has one copyright
holder, and can easily be made to run under Linux. The open synthesizers (with
the exception of ESpeak) tend to be research systems where the development
priorities aren't necessarily in line with achieving the highest speech
quality and the right feature set for applications. That is, the code is used
as a platform for experimentation and research projects rather than for the
development of applications.

SVOX Pico is a special case: the source code of the synthesizer is open, but
the tools needed to generate voices and implement new languages are not
available. These days, as I understand it, statistical techniques and machine
learning algorithms are used to derive text to speech models from collections
of linguistic data. This is how new languages can be supported quickly, for
example.