[hts-users:04234] Generate speech parameter sequences (SAT+adaptation) Error

[Marc Evrard <marc.evrard@gmail.com>]

I'm trying to synthesize speech after having trained SAT and adaptation models using the HTS2.3beta (program and demo). So I followed these steps:

– Added the lab files in the gen folder, under the required speaker folder

– rerun: make list scp

– activated these lines in the config.pm:

$MKUNG = 1; # making unseen models (GV)

$MKUN2 = 1; # making unseen models (SAT)

$PGEN4 = 1; # generate speech parameter sequences (SAT+adaptation)

$WGEN4 = 1; # synthesizing waveforms (SAT+adaptation)

– run: perl scripts/Training.pl scripts/Config.pm > log 2>&1 &

And I got this error after the $PGEN4 step:

 Components missing from Base Class list

I know that it means: some mixture components are not listed in the base class list (dec.base), although all mixture components in the model set should appear in the base class list.

But it seems that the only way to get the base class list with all the mixture is to rerun the adaptation training process. If it's the case, we need to rerun hours of training for each new utterance to synthesize, which does't sound plausible.

What am I doing wrong, or is it a bug in the demo? Could someone tell me if I follow the proper way to synthesize new utterances?

Thanks in advance for your help!

Marc