[hts-users:02062] Re: Generated speech quality

["Heiga Zen (Byung Ha CHUN)" <heiga.zen@xxxxxxxxxxxxxxxxx>]

Hi,

Girish Malkarnenkar wrote:

> I am trying to use HTS for German speech synthesis. After running the 
> cmu_arctic speaker dependent demo successfully, I replaced the raw files 
> with my own. However for creating utterance files, I had only .segment 
> files. I used dummy files for the remaining 5 files ie. 
> word,phrase,target,syllable and intevent. I was able to run training.pl 
> after a few modifications. However the final speech is very poor in 
> quality and is almost entirely unvoiced. My doubt is whether the reason 
> it is thus is because of the inadequate labelling leading to 
> insufficient information in the utt files. And if so, then can anyone 
> tell me the format of the remaining 5 files (word,phrase,target,syllable 
> and intevent) and which of them is important for the pronunciation?

Although a lot of members of hts-users and festival mailing lists are 
overlapped, I think festival's mailing list is more appropriate than 
hts-users ML to obtain answers of this question.

Note that you should also modify 
HTS-demo/data/question/question_qst001.hed because it is designed for 
ARCTIC database and doesn't match your database.

Best regards,

Heiga ZEN (Byung Ha CHUN)

--
--------------------------
Heiga ZEN (Byung Ha CHUN)
Speech Technology Group
Cambridge Research Lab
Toshiba Research Europe
phone: +44 1223 436975

______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________