[gnuspeech-contact] TRM as backend for festival

[Nickolay V. Shmyrev]

Heh, since it seems it would be hard to build Monet on Linux I've tried
to adopt trm to work with festival. Actually for me it seems it would be
interesting work, since festival predicts intonation and duration much
more precisely and is able to produce very good annotations. 

So far I've created simple voice, you can download it from 
http://festlang.berlios.de/docu/doku.php?id=gnuspeech
it is able to create some sound but it's far from perfect. So I have the
following questions:

1. What parameter should I adjust to reflect f0 changes?

2. Is the following file sounds natural for gnuspeech? Since I can't try
it on Mac in action I afraid I've missed some parameter or something was
wrong.

http://festlang.berlios.de/docu/lib/exe/fetch.php?id=gnuspeech&cache=cache&media=examples:gnuspeech-sample.wav

3. Are there any suggestion to port transitions to new engine? How
should I smooth data, should it be just a linear smooth or something
different? Also suggestions on mapping between gnuspeech and festival
radio phoneset

http://svn.berlios.de/viewcvs/festlang/trunk/festival/lib/radio_phones.scm?rev=1&view=markup

are welcome.

signature.asc
Description: Эта часть сообщения подписана цифровой подписью