[Samuel Thibault] Bug#601395: speech-dispatcher: Should have a way to use svox/pico
kyle4jesus at gmail.com
Tue Nov 2 20:41:39 CET 2010
I'm currently using SVox Pico on the latest speech-dispatcher from Git.
I'm using the native module that is included in this version rather than
the generic config file. It's working rather well so far, with the
exception of a couple of bugs which I'll list here, since as far as I
know, this is probably the best place for bugs.
Long chunks of speech are not interruptable. If I use the say all
command in Orca or page up or down through the terminal history, it
becomes impossible to interrupt speech. Only killing and restarting
speech-dispatcher fixes this problem, which I mapped to a keyboard
shortcut for testing purposes and to restart things if speech stopps or
SVox Pico seems to die if too much is going on at once, e.g. when
arrowing or tabbing through a web page or document very quickly. This
doesn't happen all the time, but it happens frequently enough to cause a
problem. Once Pico dies, there is no speech, even from dummy, and only
killing and restarting speech-dispatcher solves the problem.
The last part of a previous letter or word is spoken immediately before
the current word or letter if speech is interrupted. This may be
somewhat related to the bug I mentioned right before this one, but it
does happen all the time, even if I'm not typing quickly. Read a line of
text or type a lettor or word usingOrca and silence speech in the middle
of the utterance. Then make speech-dispatcher speak, either using Orca
or even using spd-say. a clip from the word or letter that was
interrupted will always play immediately before the new utterance is spoken.
Some letters or even whole lines are not spoken. Do an aptitude search
on an Ubuntu system from a terminal. Use orca flat review to go to the
top of the screen and attempt to read one line at a time. Pico will
speak some lines but will miss others. Also, if I try to spell a word
using orca, (double-tap keypad 5), only two to three letters in the word
are actually spoken. All other letters are ignored, and the spoken
letters seem to be in random locations, not always at the beginning of
Punctuation pauses in most cases are too long. Things like commas and
periods seem to stop speech for an unusual amount of time before the
next phrase or sentence is spoken.
To experience both the good and the bad of this rather nice synthesizer,
you will need to run speech-dispatcher from Git source, as I'm not sure
that the module is in the latest release. Also note that if you're
running Orca and want it to speak using SVox Pico, you will need to
install both Orca and speech-dispatcher in /usr/local. I tried
installing the speech-dispatcher source package over the Ubuntu
installation, and it refused to speak, and Orca seems to automatically
start /usr/bin/speechd if installed in /usr/bin, bypassing the locally
compiled version of speech-dispatcher entirely.
More information about the Speechd