[hts-users:00667] Re: question about training new voice

["Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxx>]

Hi,

Ching-Hsien Lee wrote (2007/04/24 12:11):

> after i serach hts-user ML, i quess *.utt were generated using 
> festvox (clunits voice building process), 
> right ? i want to make sure. So i can generate *.utt using my own voice 
> database.

You're right, festival-style utterance files can be generated by festvox.
However, what we really use in the training process is not festival-style utterance files, we use HTK-style context-dependent labels.
Festival's dumpfeats and utt2lab.sh are just used to convert festival-style utterance files into HTK-style labels.

So if you have already created utterance files in your own format, you can use them instead of festival-style utterance files.
In that case, what you should do is defining context-dependent label format for your target language and writing a script to convert utterance files into labels.

Regards,

Heiga ZEN (Byung Ha CHUN) 

--
------------------------------------------------
Heiga ZEN     (in Japanese pronunciation)
Byung Ha CHUN (in Korean pronunciation)

Department of Computer Science and Engineering
Nagoya Institute of Technology
Gokiso-cho, Showa-ku, Nagoya 466-8555 Japan

http://www.sp.nitech.ac.jp/~zen
------------------------------------------------