trouble with latest git, espeak, and PulseAudio

[Jacob Schmude]

Hi
Unfortunately, on this system, I do not have a choice but to use
Pulseaudio. To use ALSA directly results in audio artifacts whenever
speech is used due to poor hardware resampling in the onboard card.
Really, we have to stop saying "just ditch Pulseaudio" and work with it
instead. It's not going away, and is sometimes the only thing allowing
certain poorly-implemented hardware to work properly in cases where ALSA
and OSS4 are not enough on their own. The blanket "ditch Pulseaudio"
attitude displayed by a lot of people is worrysome. Pulse is here to
stay, to think otherwise is wishful thinking and nowhere near close to
reality.
Too bad Opensolaris doesn't support enough of the hardware in this
system for me to switch over to that, since it is free of all these
ridiculous audio layers and conflicts.


On Mon, 2010-03-01 at 14:08 +0100, Halim Sahin wrote:
> Hi,
> On Mon, Mar 01, 2010 at 03:45:28AM -0700, Jacob Schmude wrote:
> > Hi All
> > I've noticed an issue on updating to the latest git from Luke's repo.
> > Once done, I now have the issue where speech dispatcher seems to be
> > buffering too much, i.e. when reading letter by letter it will not read
> > each letter but will wait until the next letter is read, then it will
> > read both of them. However, no adjusting of AudioPulseMinLength in
> > speechd.conf is clearing this up now. It is also affecting some
> > punctuation marks in the same way, the period for example. I'm noticing
> > the same problem both when using the pulse backend and the libao backend
> > through Pulseaudio, leading me to believe it's not backend specific. Can
> > anyone else reproduce this? I'm using the espeak module, with the rate
> > cranked all the way up to 100 (equiv to 390 wpm in espeak) with the U.S
> > english voice. This happens with both Orca and Speechd-up. Turning the
> > rate down to 80 stops this from happening, but for me that is
> > torturously slow. It does seem to indicate a buffering issue though, as
> > lower rates generate more audio data and fill the buffer faster.
> > 
> > Can anyone else reproduce?
> 
> Yes sometimes. It's a known problem with pulseaudio.
> The new pulseaudio driver uses pulse like libao does.
> It adjusts tlength in pulsesimple to reduce latency.
> Sometimes pulseaudio changes it and in this case you need to restart it
> to get the lowlatency back.
> 
> If you want stable speech output, don't relay on pulseaudio.
> 
> My setup does currently the following:
> 1. I have reconfigured pulse to use alsa's dmix plugin (pa can't block
> alsa this way).
> 2. Speechd uses libao with it's alsa output driver.
> 
> HtH.
> Halim
> 
> 
> _______________________________________________
> Speechd mailing list
> Speechd at lists.freebsoft.org
> http://lists.freebsoft.org/mailman/listinfo/speechd

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20100301/e84cafb2/attachment.pgp>