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

DELIVERY FAILURE: User whchan (whchan@us.ibm.com) not listed in DominoDirectory



Your message

  Subject: Re: just a newbi problem, want to use festival, not flite with emacspeak

was not delivered to:

  whchan@us.ibm.com

because:

  User whchan (whchan@us.ibm.com) not listed in Domino Directory


Could not process part with Content-Type: message/delivery-status.

---- Begin included message ----

Just a few points which may be of interest in this thread.

Festival is a very powerful and configurable TTS. It can support quite
a few different voices and has a lot of parameters which can be
tweaked to improve synthesis. The "out of the box" configuration is a
general default that is likely to meet most users requirements.
However, I found (quite some time ago) when playing with festival, I
could get better performance tweaking some of the settings associated
with festival and the voice tools. Unfortunately, it was a while ago
and I dont actually remember what all the tweaks were.

Festival supports voices from other sources and in many distributions,
you will get different quality voices - for example, in Debian, there
are both 8k and 16k voices for US/UK english. The 8k voices don't have
the same quality, but will probably give better performance on slower
machines with less memory. 

Although I've never got around to it, mbrola voices are supposed to
provide very good quality voices with festival. Voices from
http://www.cepstral.com can also be used with festival - they had a
perl script on their website that would allow you to import a cepstral
voice into festival. Cepstral voices are good quality, but they
lose quality significantly at higher voice speaking rates. They are
commercial voices which you have to pay for, but they do have a number
of different languages available.

IMO by far the best software TTS for emacspeak is ViaVoice, which you
can purchase from http://www.wizzardsoftware.com. However, you have to
purchase the SDK and it costs over $300US. 

Second best for quality of speech is the software dectalk. However,
this TTS engine does crash regularly under emacspeak (it works
reliably under speech dispatcher, but speech dispatcher is using it in
its most basic way and does not try to use the in-line voice control
codes emacspeak uses for voice locking etc). The software dectalk is
available from http://www.fonix.com for about $100US. 

I used the softwar dectalk for quite some time and managed to live
with the regular crashes - all you need to do is a C-e C-s when the
TTS stops working and verything is restarted. Although a bit annoying,
it provides good voice quality at high speaking rates, which I think
is important. I now use it as my emergency backup TTS if I don't have
a hardware synthesiser for backup. 

HTH

Tim






Lukas Loehrer writes:
 > 
 > Some observations:
 > 
 > I just tried speech dispatcher with festival again. I am using
 > fspeech dispatcher via speechd-el this time. I am using the packages
 > from Debian unstable.
 > 
 > When using the flite output module of speech dispatcher,
 > responsiveness is the same as with eflite in emacspeak. When using the
 > festival output module in speech dispatcher, some of the problems seen
 > in espeakf go away. Speech stops immediately when it should. However,
 > there is a noticeable pause between the moment a command is issued and
 > the corresponding speech starts. I cannot tell if the pause is silence
 > generated by the tts or if it takes that long for the speech to be
 > synthesized. In the latter case, a faster machine might in fact help.
 > 
 > Thus, it seems that if one is looking for good responsiveness,
 > festival used through either espeakf or speech-dispatcher should not
 > be one's first choice. things look differently  if priorities lie
 > with support for multiple voices or languages.
 > 
 > Best regards, Lukas
 > 
 > -----------------------------------------------------------------------------
 > 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
tcross@rapttech.com.au

There are two types of people in IT - those who do not manage what they 
understand and those who do not understand what they manage.

-----------------------------------------------------------------------------
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"

---- End included message ----

Emacspeak Files | Subscribe | Unsubscribe | Search