Set of notes on tests conducted on new versions of speechd-up/speech-dispatcher

speechd-up-testing.txt (text/plain), 3.36 KB, created by Robert Grieve on 2007-06-02 12:09:40 UTC

To: [e-mail] 
From: [e-mail]
Re: Bug #179290

This bug was originally filed as speechd-up fails to start speech from rc scripts.

My thanks to the assignee for the possible fix.

I feel that this may lead to more bugs than it fixes,
but you said to let you know if it works.

In the strict sense, the answer is yes, but the answer is, not quite.

  In all the systems I have seen, a speech server is needed to carry 
  the information written to /dev/softsyn or whatever to a TTS engine. 
  This is speech-dispatcher on just about all systems I have seen. ?variants? 
  Unknown wether this is the same on all systems.

  speechd-up needs speech-dispatcher and the TTS engine to be up when it loads.
  Otherwise problems seem to abound. The depends function in speechd-up should 
  be correct be I think there's more to the problem.

As far as TTS engines, the alternatives that I know of are all listed in speech-dispatcher's 
speechd.conf in AddModule statements, along with some other ones I've never seen.

espeak-generic -- Fast,lean and very tinny, but it seems to do the trick.

Works fine on boot as long as you do:
-- emerge espeak
-- edit /etc/speech-dispacher/speechd.conf 
   set DefaultModule to 'espeak-generic'
-- Ensure alsasound is in the boot runlevel
-- do rc-update add speech-dispacher default

I think when you emerge speech-dispacher you get this one as the default

eflite -- Slightly nicer voice, a little bigger, slower

Now it works fine on boot as long as the following is done:
-- emerge eflite
-- edit /etc/speech-dispacher/speechd.conf 
   set DefaultModule to 'flite'
-- Ensure alsasound is in the boot runlevel
-- do rc-update add speech-dispacher default
-- do rc-update add eflite default (Note the 'e')

festival -- Slightly slower again, not too bad though, very nice voices.
            I wouldn't say this works on boot though, see below.
            You can set it up with

-- emerge festival
-- emerge eflite (it seems to want this for some reason -- WHY? :X
-- edit /etc/speech-dispacher/speechd.conf 
   set DefaultModule to 'festival'
-- Ensure alsasound is in the boot runlevel
-- do rc-update add speech-dispacher default
-- do rc-update add eflite default (Yes, really, it starts it anyway)

On MY system only... the following quircks still apply though.

If I have to restart the speech system for some reason, 
I would do ie

/etc/init.d/speechd-up restart

under eflite this will sometimes print 
* Starting eflite ...                                                                      [ !! ]
chown: cannot access `/tmp/es.socket': No such file or directory
chmod: cannot access `/tmp/es.socket': No such file or directory                           [ !! ]

Actually it will even do this if I have all the services stoped and do a start on speechd-up.
Doing the same command again usually suffices, but then this is still no good to a blind person.
At least it goes on boot though, and it does, so...


IMHO one uses festival at their peril.
The voices are nice, but on startup I got a very large,loud burst of white noise on my system.
Then it started working. Pass as to what that's about.

If you're out of ideas and it works reliably on boot then mark it WORKSFORME,
and I will disappear, happy.

But really a service should be able to restart without throwing a wobbly, no?
As for why festival is being that way... may the Bird of Paradise fly up its nose.