[hts-users:02793] Utterance generation

[Xavi Gonzalvo <xavi.gonzalvo@xxxxxxxxx>]

Hi,

I've been trying to generate the utterances for the SLT voice which comes with the demo in order to see if I can extract the same labels.

It looks like apart from some minor differences in the durations and F0, the main difference is the ToBi labels.

For example, first file in the demo data looks like this:

73 id _71 ; name H* ;
74 id _72 ; name H* ;
75 id _73 ; name L-L% ;
76 id _74 ; name H* ;
77 id _75 ; name H* ;
78 id _76 ; name L+H* ;
79 id _77 ; name L+H* ;
80 id _78 ; name L-L% ;

whereas I obtain this:

71 id _71 ; name L-L% ;
72 id _72 ; name L-L% ;
73 id _73 ; name H* ;
74 id _74 ; name H* ;
75 id _75 ; name H* ;
76 id _76 ; name H* ;
77 id _77 ; name L+H* ;
78 id _78 ; name L+H* ;

I've run the following:

mkdir cmu_us_slt_arctic
cd cmu_us_slt_arctic
$FESTVOXDIR/src/clustergen/setup_cg cmu us slt_arctic
./bin/do_build build_prompts
./bin/get_wavs recording/*.wav
./bin/do_build label
./bin/do_build build_utts


I'm not much into Festival and I don't really know if I'm missing any step or it's just that I'm using the wrong internal module.

Thanks

--
Xavi.