Hi,
I tried USEGV=0, the overflow problem is no longer there, but the voice quality still sounds bad comparing to the previous versions. I will keep testing the rest of the options. Sincerely, Mandy > Date: Wed, 28 Sep 2011 12:08:18 +0900 > From: uratec@xxxxxxxxxxxxxxx > Subject: [hts-users:03043] Re: Amplitude of synthesized speech while using HTS-2.2 with gv > To: hts-users@xxxxxxxxxxxxxxx > CC: uratec@xxxxxxxxxxxx > > Hi, > > Please, try other setting. > For example... > > % configure USEGV=0 > % configure USEGV=1 NOSILGV=0 CDGV=1 > % configure USEGV=1 NOSILGV=1 CDGV=0 > > Regards, > Keiichiro Oura > > > 2011/9/28 Yu-Chieh Chen <tobysworld@xxxxxxxxxxx>: > > Dear all, > > I've been trying to apply HTS-2.2 script to my own data, now i'm facing > > some amplitude issue with my training. > > I found that with > > $useGV =1; > > the amplitude of synthesized speech of some vowels is very big (overflowed) > > and makes the speech sounds broken in those phonemes. > > Is there any parameter in the training that can control this? > > Thanks in advance! > > Sincerely, > > Mandy > |