Re: Thoughts on speech

[Marco Zehe <marco zehe googlemail com>]

Hi Kenny,

I've since installed a Debian lenny distro (as you know from IRC) and  
upgraded it to sid. This is on a "real" machine, not a VM. Here are my 
observations/comments:

Kenny Hitt wrote:
> > First, I could not get the demon to start after I had installed SD. I
> > actually had to reboot the system.
> >      
> Never seen that issue, but I run all Debian boxes now days.  My recent tests with Ubuntu Hardy did have problems, but they could
> be fixed by removing a file specific to the Ubuntu speech-dispatcher package.
>    

I still had to reboot my Debian distro after installing 
Speech-Dispatcher. Even though spd-say was available and appeared to 
work, no speech came out until I actually rebooted.

> > Second, after reboot was finished, Gnome-Speech, which was still
> > configured as Orca's main speech server, was completely gone. There was
> > no speech, and since I didn't have a braille display, I was left
> > completely in the dark. I had to "blindly" start gnome-terminal, delete
> > my .orca settings folder, and shutdown and re-run Orca from the command
> > line to choose SD from the text-based initial setup
> >      
> That was because Ubuntu uses portaudio 18 for espeak.  This means espeak uses OSS for output instead of ALSA.  That issue doesn't happen
> on Debian because it uses portaudio 19 for espeak.  It will happen with any other gnome-speech synths since all the other ones use OSS.
>    

Right, this problem no longer existed on Debian.

> > Third, after I had it up and running, I found that often, two chunks of
> > stuff to be spoken would overlap, especially when Orca was speaking
> > something and I was using the keyboard to either type in something or
> > navigate. For example: I was going through the Orca Preferences dialog.
> > I had Key Echo set to Characters, and everytime I hit TAB to go to the
> > next control, it would take over half a second before Orca stopped
> > speaking the previous control that I was not interested in. Within that
> > half second, on a second audio channel, it would speak the fact that I
> > had pressed TAB. So I had both an overlap in speech output coming from
> > the SAME product, and a delay in the speech synth shutting up speaking
> > the control I was not interested in. To me, this is unnerving, if not
> > even unacceptable behaviour. If I press something like the TAB or an
> > arrow key, I want the speech engine to shut up immediately and speak the
> > new chunk of information relevant to me, and not think about whether it
> > should stop speaking and pick up the new chunk at its own leisure
> >      
> The overlap is a side effect of dmix.  I notice it, but it isn't as long here as what you experience.
>    

This is also much much better on the "real" box. Makes me believe that 
the TCP/IP traffic going back and forth between Orca and SD may have 
been influenced by the fact that this was running inside a VM, where the 
network adapters are all emulated and managed by the host operating 
system/virtualization software.

> I agree there are still issues to be resolved with speech-dispatcher, but the added features of speech-dispatcher over gnome-speech will eventually make
> Orca better once it switches to speech-dispatcher and the problems are fixed.
>    

I agree with this. But again, now that I am actually running it on a 
real box, SD leaves a much better impression with me than initially.

Marco