HMM-based Speech Synthesis System (HTS)

Content

• Intro for terminal and command execution
• Server setup
• Download packages
• Compile and set up tools
• Set up HTS demo project
• Links

Intro for terminal and command execution

If you are not sure how to execute described commands, look at Command line terminal.

Server setup

1. If tools are set up on server, some environment settings could be neccessary to set, (e.g. default locale with  UTF-8 encoding and timezone). To set up these, enter commands, e.g.:

   sudo locale-gen "en_US.UTF-8"
   sudo dpkg-reconfigure locales
   sudo dpkg-reconfigure tzdata

   where en_US.UTF-8 is necessary UTF-8 enabled locale, and proceed with setup wizards.

2. Install additionally needed packages:

   sudo apt-get install csh gcc g++ g++-multilib make libncurses5-dev libx11-dev clang byacc flex bc

Download packages

1. Create /home/HTS folder and add access permissions to all registered users to it:

   sudo mkdir /home/HTS
   sudo chmod 777 /home/HTS
   cd /home/HTS

2. Download necessary packages using wget, by executing following commands:

   wget http://festvox.org/packed/festival/2.4/speech_tools-2.4-release.tar.gz
   wget http://festvox.org/packed/festival/2.4/festival-2.4-release.tar.gz
   wget http://festvox.org/festvox-2.7/festvox-2.7.0-release.tar.gz
   
   wget --user=$WUSER --password=$WPASS http://htk.eng.cam.ac.uk/ftp/software/HTK-3.4.1.tar.gz
   wget --user=$WUSER --password=$WPASS http://htk.eng.cam.ac.uk/ftp/software/hdecode/HDecode-3.4.1.tar.gz
   
   wget http://hts.sp.nitech.ac.jp/archives/2.3/HTS-2.3_for_HTK-3.4.1.tar.bz2
   wget http://downloads.sourceforge.net/sp-tk/SPTK-3.11.tar.gz
   wget https://downloads.sourceforge.net/project/hts-engine/hts_engine%20API/hts_engine_API-1.10/hts_engine_API-1.10.tar.gz
   wget http://www.datastat.com/sysadminjournal/downloads/ActiveTcl8.4.5.0-linux-ix86-83070.tar.gz
   
   wget http://hts.sp.nitech.ac.jp/archives/2.3/HTS-demo_CMU-ARCTIC-SLT.tar.bz2

   where $WUSER and $WPASS are actual username and password received from HTK Registration form

Compile and set up tools

1. Compile speech_tools:

   tar xvf speech_tools-2.4-release.tar.gz
   cd speech_tools/
   ./configure
   make
   cd ..

2. Compile festival:

   tar xvf festival-2.4-release.tar.gz
   cd festival/
   ./configure
   make
   cd ..

3. Compile festvox:

   tar xvf festvox-2.7.0-release.tar.gz
   cd festvox/
   ./configure
   make
   cd ..

4. Extract, patch and compile HTK tools:

   tar xvf HTK-3.4.1.tar.gz
   tar xvf HDecode-3.4.1.tar.gz
   mkdir hts_patch
   mv HTS-2.3_for_HTK-3.4.1.tar.bz2 hts_patch/
   cd hts_patch/
   tar xvf HTS-2.3_for_HTK-3.4.1.tar.bz2
   cd ../htk/
   patch -p1 -d . < ../hts_patch/HTS-2.3_for_HTK-3.4.1.patch
   ./configure
   make
   sudo make install

   1. If something like following error occurs:

      esignal.c: In function 'WriteHeader': esignal.c:1184:25: error: 'ARCH'
      ...
      undeclared (first use in this function) architecture = ARCH;

   2. Add additional parameter to configure and complete compilation and installation:

      ./configure CFLAGS="-DARCH=linux"
      and compile and install HTK again.

5. Make and install hlmtools and hdecode:

   make hlmtools; sudo make install-hlmtools
   make hdecode; sudo make install-hdecode
   cd ..

   1. If compilation error occurs:

      Makefile:77: *** missing separator (did you mean TAB instead of 8 spaces?).  Stop.

   2. Open file HLMTools/Makefile and replace leading 8 spaces with tab in following line:

      ...
      mkinstalldir:
      if [ ! -d $(bindir) -a X_@TRADHTK@ = X_yes ] ; then mkdir -p $(bindir) ; fi
      ...
      and compile and install hlmtools again.

6. Extract, compile and install HTS engine:

   tar xvf hts_engine_API-1.10.tar.gz
   cd hts_engine_API-1.10/
   ./configure
   make
   sudo make install
   cd ..

7. Extract, compile and install SPTK tools:

   tar xvf SPTK-3.11.tar.gz
   cd SPTK-3.11/
   ./configure
   make
   sudo make install
   cd ..

8. Extract and install ActiveTcl (this package is probably unnecessary for latest versions of HTK):

   tar xvf ActiveTcl8.4.5.0-linux-ix86-83070.tar.gz
   cd ActiveTcl8.4.5.0-linux-ix86/
   sudo ./install.sh

   and follow setup wizard instructions.

9. Go back to HTS folder:

   cd ..

Set up HTS demo project

1. Extract and configure HTS demo project:

   tar xvf HTS-demo_CMU-ARCTIC-SLT.tar.bz2
   cd HTS-demo_CMU-ARCTIC-SLT/

2. Configure project:

   ./configure   \
   --with-tcl-search-path=/usr/local/ActiveTcl/bin \
   --with-fest-search-path=/home/HTS/festival/examples \
   --with-sptk-search-path=/usr/local/SPTK/bin \
   --with-hts-search-path=/usr/local/HTS-2.3/bin \
   --with-hts-engine-search-path=/usr/local/bin

   where /home/HTS is folder of all installed tools.

3. Make demo project:

   make

4. Run Training.pl script in background with output redirected to log file:

   /usr/bin/perl scripts/Training.pl scripts/Config.pm > log 2>&1 &

   or, if you run it on workstation, make it low priority task with command:

   nice -n20 /usr/bin/perl scripts/Training.pl scripts/Config.pm > log 2>&1 &

5. Look at the progress on log file with the command:

   tail -f log

Links

• Extraction of Speech Parameters from Speech Database using Festival