Re: [orca-list] [ANNOUNCE] libao (new audio driver for speechd including working pulseaudio support).

OK, I have a bit more feedback. I'll post here since, though I doubt Orca is at fault here, this is the most user-visible client of Speech-Dispatcher and is more likely to benefit those using it.

I've noticed that, though I have key echo enabled, some letters fail to speak when pressed, and I have to send another before they are spoken, resulting in each spoken separately. This happens always with the same letters, and never with others. In other words, there are specific characters that always trigger this behavior. It may not be instantly noticeable, as the follow-up character causes both to be spoken as if each had arrived separately. Also, if the follow-up character is on the list of characters, both are processed. So, in other words, if you send one seemingly blacklisted character then you won't get speech. Send one blacklisted character and any other, including another blacklisted char, and both are spoken separately as if each was typed separately. Sorry to be confusing, but hopefully that will make sense with a few rereads. :) Here are what appear to be the blacklisted characters for me:

a b d e g i o

I've also just noticed that this behavior occurs when reviewing the line in question char-by-char, or when backspacing over text containing them.

I wondered if perhaps the audio data for these characters was shorter than 32 bytes and maybe there was an underrun or something, so I dropped the buffer length down to 16. Still works even with a buffer size this low, but that doesn't appear to be the issue. Not sure if it happened when set to 256, but I remember noticing an occasional, odd lag, which may very well have been this issue.

On 12/17/2009 09:23 AM, Marco Skambraks wrote:
cool feedback - thanks
if you want to try different settings for latency etc.
you can change the define in speechd/src/audio/libao.c
#define AO_SEND_BYTES 256

if you reduce this value 128,64,32 ... and re-compile speechd
you can generate a faster resonse play/stop

the reason for me to create the patch was the alsa echo problem
if speech-dispatcher gets new strings very rapidly
the sound-output was a kind of a mixed speech output
sounds lke an echo
with libao and alsa there is no echo anymore :-)

happy testing


On Thu, 17 Dec 2009, Hammer Attila wrote:

Dear List,

Halim, this driver is fantastic my machine.
I tested the new driver.
Before with this testing driver is publicated, Pulseaudio doing 2-3 second latency when Speech-dispatcher is running in Ubuntu Karmic my notebook sound card and my desktop computer sound card.

I need remove Pulseaudio and coming back with alsa and do lot of workarounds to solving problems (volume increase/decrease workaround with custom keybinding command with amixer, remove unneed pulseaudio packages if I would like listen musics, videos etc).

Now when I testing the new driver, I am try the driver with Ubuntu Lucid system.

What can I see after when all required steps I doing the driver installation? I get back my fast speech navigation when Pulseaudio is running my notebook. Only a very minimal latency I see when I typing, but this is not matter (not 3 second latency). :-):-) Alsa is possible very little faster. But the new testing driver is very stable. Tomorrow I testing longest time.

What can I do when I would like compile and install the new driver?
Halim and me wroted 2 or 3 priwate letter, final, following simple method is working my machine. Note: Please, ewerybody try this method with only your risk! Important, following method working with Karmic and Lucid only. Halim, if I any step writing incorrect, please complete. Excuse me my bad english:

1. In Lucid, I downloaded speech-dispatcher source code from following command with a directory:
sudo apt-get source speech-dispatcher

2. Please do following commands:
sudo apt-get install libao-dev (this is important)
sudo apt-get build-dep speech-dispatcher

3. I navigate the speech-dispatcher source directory.

4. Applyed the new patch:
patch -p1 <../speechd-libao.diff
Please replace the patch path if you storing Halim sent diff file with another place. :-):-)

5. I run dpkg-buildpackage command.

6. I installed with following packages with prewious directory (need do a sudo cd .. command):
speech-dispatcher, python-speechd, libspeechd2.
Possible your system needs another packages, but you found with rebuilded packages with your needs. Possible the version number is change with your system.

7. I change the /etc/libao.conf with default_driver=alsa to default_driver=pulse.

8. Final, reconfiguring speech-dispatcher with using libao audio output module. Don't worry, my machine spd-conf ask me the suggested way is Alsa, Pulse, Oss, but don't need correct your answer. :-):-)

Enjoy, the new driver working my machine.

Orca-list mailing list
Orca-list gnome org
Visit for more information on Orca.
The manual is at
The FAQ is at
Netiquette Guidelines are at
Log bugs and feature requests at

The OpenBlinux Project
Orca-list mailing list
Orca-list gnome org
Visit for more information on Orca.
The manual is at
The FAQ is at
Netiquette Guidelines are at
Log bugs and feature requests at

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]