[orca-list] Espeak variants was Re: New Linux user, needing higher quality speech

[Luke Yelavich <themuso ubuntu com>]

On Mon, May 18, 2015 at 11:26:57AM AEST, Luke Yelavich wrote:
> Re-sending to the Orca list, forgot to reply to all.
>
> On Mon, May 18, 2015 at 09:16:28AM AEST, kendell clark wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA512
> >
> > hi
> > Myself, I think this could be partially solved by implementing proper
> > varient support for speech-dispatcher's espeak module. The nvda espeak
> > driver does this, but it does it, I believe, by having it's own
> > special copy of espeak, and this would be all kinds of inconvenient to
> > deal with, so if this is done, It should probably either be done by
> > interfacing with libespeak, or somehow communicating with the espeak
> > code itself via system calls.
>
> Speech Dispatcher's espeak driver already uses libespeak to do all speech synthesis. I think at the time 
> the espeak driver was developed, Espeak didn't have variants. It also doesn't help that Speech Dispatcher 
> has the concept of variants, but in Speech Dispatcher terms, variants referrs to languages.
>
> Espeak also doesn't seem to list its variants in its voice list output, either via Speech Dispatcher or the 
> espeak command-line utility. So perhaps espeak does need to be extended to make its variants available via 
> the same way other voices are listed and used.

Ok, after some digging, it is possible for espeak, and even Speech Dispatcher to set a voice with a variant. 
In other words, if a Speech Dispatcher client is using espeak and knows a voice variant name, it can set the 
variant along with a voice directly. Speech Dispatcher's espeak driver itself makes use of the variants for 
the pre-defined male and female voices in the Speech Dispatcher API.

Because variants are meant to be used with all other voice files, trying to list all voices with all variants 
makes for a very long voice list, which is probably why espeak doesn't currently do this. We could check for 
a list of variants from the Speech Dispatcher driver, but again it makes for a long voice list, and even if 
we did that, Speech Dispatcher doesn't provide anything generic for clients to set things such as the voice 
variant. I do have something in mind to allow for synthesizer specific settings however.

Luke