[hts-users:04202] Re: Creating *.lab and .utt files

["Toman, Markus" <Toman@ftw.at>]

Hi,

as a starting point you can use the HTS demo "HTS-demo_CMU-ARCTIC-SLT.tar.bz2" from
http://hts.sp.nitech.ac.jp/hts-users/spool/2014/msg00201.html

Basically just replace the utts/labs and the raw sound files with your data and run it to train a model.

You can then use the resulting .htsvoice model file and a label file with hts_engine to synthesize wavs (e.g. using the command line tool).

At that point you can also consider using flite to synthesize from text. You can also use our SALB framework (https://github.com/m-toman/SALB) which includes flite for English but also can read festival lexicon/letter-to-sound-files for synthesis of other languages (it also includes everything you need to build a SAPI5 voice for Windows).

Best
Markus


--
Dipl.-Ing. Markus Toman
Researcher
FTW Telecommunications Research Center Vienna
Donau-City-Straße 1, A-1220 Vienna, Austria

--

From: Afruza Begum [mailto:afruza.moon@gmail.com] 
Sent: Donnerstag, 22. Jänner 2015 05:52
To: hts-users@sp.nitech.ac.jp
Subject: [hts-users:04201] Re: Creating *.lab and .utt files

help
I create .utt and .lab files by using festival and i already build context dependent HMM model in HTK. Now how i can use .utt and .lab file to produce synthesis speech in HTK. Please  help me if anybody knows.

On Tue, Jan 20, 2015 at 5:34 PM, Sébastien Le Maguer <slemaguer@coli.uni-saarland.de> wrote:
Hello,

based on the forced aligment documentation on the HTKBook (page 208), results you've got at the model level (using the -m option) are respecting this format

<start> <end> <phone_label> <score> <word_label>

Therefore you need to use the 3 first parts (start, end and phone_label) to get the monophone label files. For the full label files, you need to add additionnal properties of "your" choices to the phone_label. Don't forget that you need to provide an associated question file consistent with the property values used in the full label files.

The utt files are obtained using the festival TTS system. If you don't use it, you don't need it as long as you get the label files in other way.

Kind regards,
Sébastien Le Maguer

Kumara Guru writes:

> help
>
> I am trying to do speech synthesis for a new language.
> Is there a step by step instructions listed somewhere ...
> (like in case of HTKBook)
>
> I am able to generate force aligned files using htk.. Not able to figure out
> how to
> get HTS *.lab and *.utt files..
>
> guide
--
Dr. Sébastien Le Maguer
Postdoctorate researcher

Saarland University
Campus C7.4 - room 2.03
D-66123 Saarbrücken
Germany

phone : +49-681-302-70030
Mail: slemaguer@coli.uni-saarland.de
website : http://sites.google.com/site/seblemaguer