[Kde-accessibility] punctuation or other improvments?

[david powell]

On Sunday 28 October 2007 02:24:18 Jonathan Duddington wrote:
> On 28 Oct, Gary Kline <kline at ethos.thought.org> wrote:
> > 	Then perhaps you can tell me the relationship between Alan Black's
> > "Festival" and "espeak";
>
> There is none.
> eSpeak is a different open-source text to speech synthesizer.  It uses
> a different method, "formant synthesis".  This makes it faster (more
> responsive) and much smaller (which allows several languages in a small
> package), but less natural in speech quality.  However I believe its
> enunciation is clearer, which was your main problem.
>
> > there is a suite of three speech utilities named "*mbrola" that have
> > not up updated on my version of BSD for two years.   I would like to
> > use the leading synthsizer--the one most actively being developed.
>
> The mbrola diphone voices are only the back-end part of a synthesizer.
> They need a front-end to do spelling to phoneme conversion and
> intonation.  eSpeak can be used for this and may be useful to avoid the
> Festival errors which you mentioned, while avoiding the mechanical
> sound of eSpeak's own formant synthesis.
>
> Basically I'm suggesting that you try alternatives to find which sounds
> best for you.  Different people have different preferences.
>
> > I'll send you some words, off-list.
>
> Thanks.  It's easy enough for me to find words which are pronounced
> wrongly, by speaking a lexicon.  But many are uncommon words, or are
> only slightly wrong, or both.  Many may not be a problem in practice.
> But if you find errors which are annoying or confusing, then please
> tell me.
>
> _______________________________________________
> kde-accessibility mailing list
> kde-accessibility at kde.org
> https://mail.kde.org/mailman/listinfo/kde-accessibility

with the festival based voices you will find that theres a dictionary file for 
it in  /usr/share/festival/dicts/<the name of the voice>/
although it mybe a bit involved you can define how each word is spoken by 
defaut there as well as add new words 

quite often you have to trick it into pronouncing the word correctly 
for example  for "email"  put  "he mail" as the text you want it to speak

theres one or two quirks in the standard festival dictionaries that i find 
annoiying and have corrected in the dictionary files  like DEC is spoken 
as "december" rather than decriment, short for digital equipment corp 
or many other abreviations also 

unfortunalty form thepoint of view of kdeaccessibilaty and kttsd its not realy 
posible to have patches for festival in the install package both from the 
licencing point of view and the software side of installing files inot 
another package 

what may be a solution is to use the kttsd filters section and create your own 
filters for the words you want it to speak diferently to standard 

realy you should be using festival 1.95 beta and one of the arctic database 
based voices http://festvox.org/cmu_arctic/dbs_slt.html
for best results use the multisyn versions of the voice

or if your realy brave you can rebuild the voice form the original source 
audio files to get better audio quality also (if you have a few months to 
spare that is )

the festival mailing list is  
festival-talk at festvox.org 


David (aka achiestdragon)
personal website http://www.achiestdragon.org
artwork http://achiestdragon.deviantart.com/