Software Synths

[hanke at volny.cz]

On Mon, Jan 10, 2005 at 09:01:42AM -0500, Janina Sajka wrote:
> You haven't found documentation, because no one has written any. Perhaps
> you'll fix that for the next person if you see this through.

Hello everyone,

I'd like to clarify that a bit, because the problem here is more
serious than missing documentation. We haven't written any documentation
about using Speakup with software speech because there is no clear
way how to do that properly. That's also the reason why we didn't
produce any distribution packages for speechd-up.

I personally don't recommend speechd-up for serious work. I'm aware
of many issues with it (slow response on typing keys, impossible to
use with any other language than English, impossible to modify speech
parameters from outside, punctuation and capital letter recognition
modes not working properly, ...). 

I think the issues with slow responsibility are the worst ones. Not even
it is annoying, but it leads many users to think software speech on
today's computers is very slow and inferior to hardware synthesis. This
is not true, that's a problem in the interface Speakup--Speech
Dispatcher.

I don't see any possible way to fix speechd-up under the current
interface provided by Speakup in /dev/synth. If we would like to
interface Speakup to software speech (using Speech Dispatcher or
other methods) properly, that would require (some trivial and some
non-trivial) modifications to Speakup.

We tried to do them in Brailcom, but we didn't succeed. Partly because
we didn't have any previous experience with kernel programming and
partly because Speakup's code is quite complicated and not well
documented (as well as the kernel code itself in this area).

If there is will among the Speakup developers to try to solve this
problem (I'm not speaking about any technical details right now),
we really welcome it and we are ready to cooperate.

But till then, or till somebody develops some other decent screen
reader (I would especially welcome an effort to develop a user-space
screen reader), I can't recommend using speechd-up unless there is
no other choice.

I'd especially like to bring the speechd-el and Emacspeak approach
to your attention as a possible partial solution. You can find out
more about speechd-el on http://www.freebsoft.org/speechd-el
and about Emacspeak on http://emacspeak.sourceforge.net . I think this
approach is much better in many cases. On the other side it can't substitute
Speakup in providing speechd during boot/shutdown and when you run
into problematic situations.

With Regards,
Hynek Hanke
http://www.freebsoft.org/