Re: ViaVoice Outloud, PulseAudio and Ubuntu 11.04

["Tim Cross" <tcross@xxxxxxxxxxx>]

Hi Bart,


I notice no latency issues with viavoice. However, I've never experienced
latency issues with viavoice. The only issue I've ever encountered with
viavoice has been distored sound, extremely slow (i.e. 1 word per hour - at
least that is how it seemed) and choppiness. This I normally had to solve by
playing around with ALSA settings in .asoundrc. Once I had it working well with 
ALSA, the quality and responsiveness have always been excellent. I notice no
difference in latency between via voice with just ALSA or viavoice going
through pulseaudio via the alsa plugin. 

The ViaVoice server has been the only server I've used which is able to keep up
with my typing and is responsive enough to still be intelligible when moving
around buffers quickly etc. The espeak server on the other hand is a different
story. Even when I've got it working pretty well in that it does not chop of
words and responds quite quickly, it cannot keep up with my typing rate and I
have to turn of character echo. It is also sluggish in responding when moving
around buffers etc. However, I think some of this is due to the way the
emacspeak espeak server is written. It does a lot of re-setting of 'meta' data
as part of its support for multiple languages etc and I don't think most of it
is necessary. I hacked a version on my home machine and it appears to run much
better than the one I'm running here at work, which is stock standard. I need
to look at this more though as there are other significant difference. For
example, the home system is 64 bit. 

I just downloaded the latest dev version of espeak and built it with pulseaudio
support. While it is working, its latency is really really bad. In fact it
won't even keep up with word echo when typing! I was trying to use it while
writing this email and had to switch back to outloud as it was beginning to
block emacs. Again, I think this is something to do with the tclespeak.so
library. I will have to compare it with my hacked one at home. It could also be
an issue with the latest espeak version as it seemed to be working better with
the previous stable version. 

I just played around with viavoice a bit and to be honest, cannot identify any
latency issues at all. Character echo is able to keep up with my fast typing
rate. As I scroll through this file line-by-line, I note that even when moving
from line to line quickly, it neatly chops off the previous line and starts
speaking the new line almost instantly i.e. I cannot 'arrow up' quickly and
have a heap of speech events queue up and then have to wait for them to clear
as happens with espeak. 

Some time ago, I said I was going to start doing some low level documentation
and analysis of the emacspeak speech server libraries etc. This was mainly so
that I got to understand it enough so that I could perhaps improve the espeak
server. Unfortunately, I got sidetracked doing quite a bit of development and
bug fixing for the VM project. However, I've pretty much finished what I wanted
to do there, so now may be a good time to get back to this. 

Tim

 
Bart Bunting writes:
 > Hi Tim,
 > 
 > Can you make any comments regarding the latency of the viavoice server
 > with this setup?
 > 
 > Cheers
 > 
 > Bart
 >  
 > On Wed, 18 May 2011 14:50:16 +1000, "Tim Cross" <tcross@xxxxxxxxxxx> wrote:
 > > 
 > > I'll preface the following by saying that this is only an outline of my recent
 > > experiences and others may not have as much luck. Issues, such as sound card
 > > type and manufacturer, ALSA driver support etc, means there can be a lot of
 > > variation in this area. 
 > > 
 > > A while back, I posted to this list regarding the success I had achieved with
 > > espeak and pulse audio. I reported that I'd found that the emacspeak espeak
 > > driver, compiled with portaudio support did not work well. It was sluggish and
 > > had a tendency to chop off ends of words or skip words. I found that
 > > re-compiling it against pulseaudio rather than portaudio fixed these issues. 
 > > 
 > > At the time, despite numerous different strategies, I was never able to get
 > > good quality ViaVoice outloud output with pulse audio. In fact, I had never
 > > been able to get good quality outloud output without customizing ALSA via a
 > > .asoundrc file.  
 > > 
 > > Recently, I have upgraded my system to Ubuntu 11.04. I decided to re-examine
 > > matters and see if I could get a better outcome. One problem I had with the
 > > existing setup was that when running emacspeak and the outloud server, pulse
 > > audio was blocked and so I could not run any app built with pulse audio. 
 > > 
 > > Given this was an upgrade and had lots of new libraries, my first step was to
 > > remove all alsa and pulse audio configurations. I removed my .asoundrc file and
 > > restored all pulse settings to their default, rebooted the system and then
 > > started playing around to see what I needed to do to get as much working as I
 > > could. 
 > > 
 > > The outcome was surprising. For the first time ever, all my sound related
 > > systems on this desktop are working 'out of the box'. I have Outloud running
 > > via pulse audio, can run multiple sound sources, both pure ALSA and pulse audio
 > > systems simultaneously and have full control over each independently of each
 > > other. 
 > > 
 > > The system I am using is a Dell 32bit desktop with an internal 'on-board'
 > > soundcard from Intel. I'm running Ubuntu 11.04. This system has been running
 > > Ubuntu since 8.04 and has only been upgraded and not re-installed since its
 > > original setup. 
 > > 
 > > Looks like Linux sound support is really beginning to mature and provide
 > > similar functionality to what is common on other more expensive and locked down
 > > solutions. 
 > > 
 > > Tim
 > >  
 > > -- 
 > > Tim Cross
 > > Information Technology 
 > > University of New England
 > >  Phone: +61 2 6773 3210
 > > Mobile: 0428 212217
 > >    Fax: +61 2 6773 3424
 > > E-Mail: tcross@xxxxxxxxxxx
 > >    Web: http://www.une.edu.au/itd
 > > ---
 > > Please avoid sending me Word or PowerPoint attachments.
 > > See http://www.gnu.org/philosophy/no-word-attachments.html
 > > 
 > > Any fool can make things bigger, more complex, and more violent. It takes a
 > > touch of genius (and a lot of courage) to move in the opposite direction.
 > >                                                            âAlbert Einstein
 > > 
 > > 
 > > 
 > > 
 > > 
 > > 
 > > -----------------------------------------------------------------------------
 > > To unsubscribe from the emacspeak list or change your address on the
 > > emacspeak list send mail to "emacspeak-request@xxxxxxxxxxx" with a
 > > subject of "unsubscribe" or "help".
 > > 

-- 
Tim Cross
Information Technology 
University of New England
 Phone: +61 2 6773 3210
Mobile: 0428 212217
   Fax: +61 2 6773 3424
E-Mail: tcross@xxxxxxxxxxx
   Web: http://www.une.edu.au/itd
---
Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html

Any fool can make things bigger, more complex, and more violent. It takes a
touch of genius (and a lot of courage) to move in the opposite direction.
                                                           âAlbert Einstein






-----------------------------------------------------------------------------
To unsubscribe from the emacspeak list or change your address on the
emacspeak list send mail to "emacspeak-request@xxxxxxxxxxx" with a
subject of "unsubscribe" or "help".