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

Re: Comparing the responsiveness of espeak and Outloud

Hi Alastair,

its interesting that you find espeak more responsive than outloud. I
have found almost exactly the opposite. While I don't notice any real
difference in stopping - at elast nothing I've ever noticed, I do
notice a bit difference in responsiveness with respect to typing and

With outloud, I have both character and word echo enabled. Outloud has
no problems keeping up with my typing, with is somewhat inaccurate,
but quite fast. Espeak on the other hand is unable to keep up. In
fact, it is so slow, I have to turn off character echo and only use
word echo. Even then, I find it can stil get a little behind.

I do use espeak compiled against pulseaudio rather than portaudio
becuase I found it had lots of problems when using portaudio -
chopping off utterances, repeating bits of utterances etc. Once I
recompiled against pulsaudio, it worked well, but was not as
responsive as I would like. I have 'trimmed' down the espeak code a
bit and have been experimenting to try and see if I can find a better
setup. For example, espeak uses SSML markup to change voices and
espeak provides separate functions for speaking a key, a wide (UTF-8)
character and speaking text. I'm trying to determine whether using
speakCharacter or speakKey or speakText is best or if plain text is
better than SSML marked up text or if some other embeded coding and
use of setPitch and setPitchRange would work better etc. I'm also
experimenting with using Java/Clojure as the implementaiton langauge
and using other synthesizes, such as the Mary TTS. If I get time, I
also plan to do an interface to speech-dispatcher.

Not a lot of this is much help to you and you probably won't get a lot
of help because there are few who have the level of familiarity with
the various servers than you would think. While Raman knows the
outloud and dtk-exp servers well, I think others did the owrk for
esepak, mac etc. In the end, unfortunately, to answer your questions,
you are probably going to have to do the research and climb the step
knolwedge path yourself. We need more people to contribute in this
area, so any information you can dig up or the resoults of any testing
you do are always welcome.


On 28 February 2012 14:08, T. V. Raman <tv.raman.tv@gmail.com> wrote:
> This is useful feedback. As to why the emacspeak server doesn't
> use that call, simple answer may be that that call didn't exist
> at the time Iimplemented the emacspeak alsa server -- from memory
> sometime in 2003. I'll look at what that function does when I
> have some time.
>>>>>> "Alastair" == Alastair Irving <alastair.irving@sjc.ox.ac.uk> writes:
>    Alastair> Further to the below, I've now got voxin working
>    Alastair> with speakup via the speakup connector. Stopping
>    Alastair> speech with it is far more responsive than with
>    Alastair> emacspeak.
>    Alastair>
>    Alastair> I've had a look at the code for both the emacspeak
>    Alastair> server and the speakup connector. The main
>    Alastair> difference seems to be that the connector uses the
>    Alastair> snd_pcm_avail_update method to determine how many
>    Alastair> frames to write. It then doesn't do any checking
>    Alastair> for underruns.
>    Alastair>
>    Alastair> I don't really understand all the details of what's
>    Alastair> happening but is it possible that this explains the
>    Alastair> difference in performance? If so is there a reason
>    Alastair> why the emacspeak server doesn't use the
>    Alastair> snd_pcm_avail_update method?
>    Alastair>
>    Alastair> Alastair
>    Alastair>
>    Alastair> On 20/02/2012 18:35, Alastair Irving wrote:
>    >> Hi
>    >>
>    >> I'm running Arch Linux inside a virtual machine. I have
>    >> emacspeak 34.0 with both the espeak and Outloud servers
>    >> compiled and working. Both servers work well, the only
>    >> problem is that the outloud server seems to be less
>    >> responsive than espeak, especially when trying to stop
>    >> speech. It's often the case that when stopping outloud an
>    >> extra couple of sylabuls will be spoken. I know this
>    >> doesn't sound like much but if you cursor through lines
>    >> very quickly it can get to be a long way behind.
>    >>
>    >> I did some experimenting by running the outloud server
>    >> directly, queueing some speech with the q command,
>    >> speaking it with d and stopping it with s. as above there
>    >> was a very obvious delay in stopping, where as with espeak
>    >> there is not. Whilst doing this I realised that every time
>    >> I got it to speak something the server printed numerous
>    >> messages of the form Underrun!!! (at least ... ms too
>    >> long) (for varying values of ...).
>    >>
>    >> Presumably this shouldn't happen.
>    >>
>    >> I would really appreciate it if someone who understands
>    >> the workings of the server could please tell me if the
>    >> slow stopping and the underrun issues are the same problem
>    >> or different and if there is anything I can do about the
>    >> slow stopping problem.
>    >>
>    >> Many thanks
>    >>
>    >> Alastair
>    >>
>    >> -----------------------------------------------------------------------------
>    >> To unsubscribe from the emacspeak list or change your
>    >> address on the emacspeak list send mail to
>    >> "emacspeak-request@cs.vassar.edu" with a subject of
>    >> "unsubscribe" or "help".
>    >>
>    Alastair>
>    Alastair> -----------------------------------------------------------------------------
>    Alastair> To unsubscribe from the emacspeak list or change
>    Alastair> your address on the emacspeak list send mail to
>    Alastair> "emacspeak-request@cs.vassar.edu" with a subject of
>    Alastair> "unsubscribe" or "help".
> --
> Best Regards,
> --raman
> -----------------------------------------------------------------------------
> To unsubscribe from the emacspeak list or change your address on the
> emacspeak list send mail to "emacspeak-request@cs.vassar.edu" with a
> subject of "unsubscribe" or "help".

Tim Cross

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

If you have questions about this archive or had problems using it, please send mail to:

priestdo@cs.vassar.edu No Soliciting!

Emacspeak List Archive | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001 | 2000 | 1999 | 1998 | Pre 1998

Emacspeak Files | Emacspeak Blog | Search the archive