16k voices

Jacob Schmude jschmude at adelphia.net
Thu Sep 18 04:24:08 EDT 2003


Hi
Honestly, I'd recommend a hardware solution at this point. There are 16k voices out there for festival, flite and freetts. However, that doesn't necessarily 
have anything at all to do with the quality of the voice. Festival, flite, and freetts (software synthesizers) all use the same basic voices. These voices are, in 
my opinion, terrible. Oh, they're clear due to there being 16K, but their inflection is almost nonexistant and, due to the pronunciation they use, they tend to be 
difficult to understand. You can get the software synths and voices at:
festival: http://www.festvox.org/
flite (festival lite): http://cmuflite.org
freetts (java-based speech engine) http://freetts.sf.net/
As of right now, festival is the only software synthesizer supposedly supported by speakup 1.5. I say supposedly because I've never gotten it working 
satisfactorily. You need to apply a patch and run an add-on daemon, but it's incredibly slow to respond, doesn't allow changing of voice parameters like rate 
and pitch, and hasn't been updated or changed since april according to the filename. You can get the patch at:
http://users.wpi.edu/~blinux
if you want to try it.
IMHO, the only good linux software synth is software dectalk, but that costs $50 and speakup doesn't work with it yet. Hint, hint, speakup developers? Any 
chance it'll ever be supported?
If you need help getting the software synthesizer working, let me know. It can be a bit of work, especially if you're going to try festival. Software Speech is 
one of the two major areas in which linux lags behind, I think, the other being OCR.
HTH
On Wed, 17 Sep 2003 23:12:47 -0400, Sina Bahram wrote:

>Hello, I read a few earlier posts on this and other lists about some 16k
>voices availible for linux. I was wondering what would be availible, and
>if 16k voices are truly the highest quality availible these days.







More information about the Speakup mailing list