[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:02672] Re: synthesizing waveforms using hts_engine



Hi Alexis.
thanks for reply. i understand what you say. i need to use hts_engine with different input (*.lab) files to synthesize another sentence.
do you know about *.lab and *.trace files? how can i create *.trace file for different sentences?

On Wed, Jan 5, 2011 at 2:46 PM, Alexis Moinet <alexis.moinet@xxxxxxxxxxx> wrote:
Hi Ramil,

In the hts-demo, you can add a 'print' in the file scripts/Training.pl in the section of code that synthesizes speech with hts_engine to log the exact command line that is used to run hts_engine

in hts-demo 2.1.1 it is in the section at lines 681-731 (surrounded by if ($ENGIN) { ... } )

change this :

print "Synthesizing a speech waveform from $lab using hts_engine...";
shell("$hts_engine -or ${dir}/${base}.raw -ot ${dir}/${base}.trace $lab");

to this :

print "Synthesizing a speech waveform from $lab using hts_engine...\n";
print "$hts_engine -or ${dir}/${base}.raw -ot ${dir}/${base}.trace $lab";
shell("$hts_engine -or ${dir}/${base}.raw -ot ${dir}/${base}.trace $lab");

and run the demo (set all switches but $ENGIN to 0 in Config.pm if you've already run the demo beforehand)

if you read the log file you should see something like this :

/usr/local/bin/hts_engine -td ./hts-demo/voices/qst001/ver1/tree-dur.inf -tf ./hts-demo/voices/qst001/ver1/tree-lf0.inf -tm ./hts-demo/voices/qst001/ver1/tree-mgc.inf -md ./hts-demo/voices/qst001/ver1/dur.pdf -mf ./hts-demo/voices/qst001/ver1/lf0.pdf -mm ./hts-demo/voices/qst001/ver1/mgc.pdf -dm ./hts-demo/voices/qst001/ver1/mgc.win1 -dm ./hts-demo/voices/qst001/ver1/mgc.win2 -dm ./hts-demo/voices/qst001/ver1/mgc.win3 -df ./hts-demo/voices/qst001/ver1/lf0.win1 -df ./hts-demo/voices/qst001/ver1/lf0.win2 -df ./hts-demo/voices/qst001/ver1/lf0.win3 -s 16000 -p 80 -a 0.42 -g 0 -l -b 0.4 -cm ./hts-demo/voices/qst001/ver1/gv-mgc.pdf -cf ./hts-demo/voices/qst001/ver1/gv-lf0.pdf -b 0.0  -or ./hts-demo/gen/qst001/ver1/hts_engine/alice01.raw -ot ./hts-demo/gen/qst001/ver1/hts_engine/alice01.trace ./hts-demo/data/labels/gen/alice01.lab

you can run this as a command line, you just need to use different input (*.lab) and output files (*.raw and *.trace) in order to synthesize another sentence (i.e. another *.lab than alice01.lab)

running :

hts_engine -h

in command line will give you some clues about which option of hts_engine does what.

NB: in Training.pl, the output speech file (*.raw, 16bits/PCM) is then rewrittent as a *.wav file using sox (line 727).

Hope it helps,

Alexis

Ramil wrote :
hello everyone.
i am using HTS-demo_CMU-ARCTIC-SLT speakerdependent. i created my modells successfully. and my /hts/HTS-ARCTIC-AZ/gen/qst001/ver1/hts_engine test files sounds well.
now, i want to synthesizing waveforms using hts_engine. i googled, but nothing found.
can anyone help me? how can i synthesizing waveforms using hts_engine?


--
Best Regards,
        Ramil Mammadov



--
Best Regards,
        Ramil Mammadov

Follow-Ups
[hts-users:02673] Re: synthesizing waveforms using hts_engine, Alexis Moinet
References
[hts-users:02670] synthesizing waveforms using hts_engine, Ramil
[hts-users:02671] Re: synthesizing waveforms using hts_engine, Alexis Moinet