festvoxtofreetts.sh

FreeTTS is a speech synthesis system written entirely in the Java programming language. It is based

#!/bin/bash
# Portions Copyright 2003 Sun Microsystems, Inc.
# Portions Copyright 1999-2003 Language Technologies Institute,
# Carnegie Mellon University.
# All Rights Reserved.  Use is subject to license terms.
#
# See the file "license.terms" for information on usage and
# redistribution of this file, and for a DISCLAIMER OF ALL
# WARRANTIES.

usage() {
    echo "Usage: $0 <voicedir> [lpc | sts | mcep | idx | install | compile]"
    echo "       $0 --help"
    echo "Converts a festvox voice into FreeTTS format."
    echo
    echo "--help            Show this message."
    echo "<voicedir>        The directory containing the festvox voice data."
    echo "                    This directory must have a wav/ subdirectory."
    echo "lpc,sts,mcep,...  The second parameter may explicitly run an"
    echo "                    individual stage of the conversion process."
    echo
    echo "The ESTDIR environment variable must point to the directory"
    echo "  containing a compiled version of the Edinbourough Speech Tools."
    echo "festival, ant, java, and javac must be in your path."
    echo
    echo "Running with no second parameter will run the stages in order:"
    echo "  (lpc,sts,mcep,idx,install,compile)."
    echo "Note that some stages may rely on the execution of previous stages."
    echo
    echo "It is recommended to execute the conversion process without"
    echo "  the second parameter."
    exit -1
}

if [ "$1" = "--help" ] || [ "$1" = "-h" ] || [ "$1" = "-help" ]; then
    usage
fi

if [ -d "$1" ] && [ -d "$1/wav" ]; then
    VOICEDIR="$1"
else
    usage
fi

. $VOICEDIR/etc/voice.defs

# Description of freetts-specific properties. Try to read them
# from voice dir, or assume default values. This is only needed
# for the "install" and "compile" steps.
if [ -e "$VOICEDIR/etc/freetts.properties" ]; then
  . $VOICEDIR/etc/freetts.properties
else
  # defaults
  VP_NAME="$FV_NAME"
  VP_GENDER="NEUTRAL"
  VP_AGE="NEUTRAL"
  VP_DESCRIPTION="description not available"
  VP_FULL_NAME="$FV_VOICENAME"
  VP_LOCALE="en_US"
  LOCALEPATH="en/us"
  VOICETARGETBASE="com/sun/speech/freetts"
fi

if [ "$2" = "compile" ]; then
    if [ "$FV_TYPE" = "diphone" ]; then
        ant -Ddiphone_voice=$VP_FULL_NAME -Duser_voice_base_path=$VOICETARGETBASE/$LOCALEPATH -find build.xml
    elif [ "$FV_TYPE" = "ldom" ] || [ "$FV_TYPE" = "clunits" ]; then
        ant -Dclunit_voice=$VP_FULL_NAME -Duser_voice_base_path=$VOICETARGETBASE/$LOCALEPATH -find build.xml
    fi
    exit 0
fi

if [ ! "$ESTDIR" ]; then
    echo "environment variable ESTDIR is unset"
    echo "set it to your local speech tools directory e.g."
    echo '   bash$ export ESTDIR=/home/<username>/projects/speech_tools/'
    echo or
    echo '   csh% setenv ESTDIR /home/<username>/projects/speech_tools/'
    echo
    usage
fi

if [ ! -f $VOICEDIR/etc/voice.defs ]; then
   echo "Can't find $VOICEDIR/etc/voice.defs file"
   echo "don't know what voice to convert"
   echo
   echo "If the voice directory is correct, you could try"
   echo " festvox/src/general/guess_voice_defs to generate it."
   echo
   echo "Setup for conversion of $1 to flite FAILED"
   echo
   usage
fi

if ! festival --version; then
    echo
    echo "ERROR: festival not in path."
    echo
    usage
fi

if ! java -version >/dev/null 2>/dev/null || ! javac -help 2>/dev/null; then
    echo
    echo "ERROR: java and javac must be in path."
    echo
    usage
fi

if ! ant -version; then
    echo
    echo "ERROR: ant not in path."
    echo
    usage
fi


# perform conversion
if [ "$2" = "" ]; then
    
    # perform each step individually
    if ! $0 $VOICEDIR lpc; then
        exit $?
    fi
    if ! $0 $VOICEDIR sts; then
        exit $?
    fi
    if [ "$FV_TYPE" != "diphone" ]
    then
        if ! $0 $VOICEDIR mcep; then
            exit $?
        fi
    fi
    if ! $0 $VOICEDIR idx; then
        exit $?
    fi

    if ! $0 $VOICEDIR "install"; then
        exit $?
    fi

    if ! $0 $VOICEDIR compile; then
        exit $?
    fi

    echo
    echo "Conversion process complete"
fi

# The scheme and java files should be in the same directory as this script
HELPERDIR=`dirname $0`
# Make sure that HELPERDIR contains an absolute path:
echo $HELPERDIR | grep "^/" > /dev/null || HELPERDIR=`pwd`/$HELPERDIR

# This assumes that FreeTTS is configured with this directory structure
FREETTSDIR="$HELPERDIR/../.."

# We need some files from the ArcticToFreeTTS directory...
ARCTICDIR="$FREETTSDIR/tools/ArcticToFreeTTS"

#This is where some temperary files are generated as well as the final voice
OUTDIR=$VOICEDIR/FreeTTS
mkdir $OUTDIR >/dev/null 2>/dev/null

(cd $ARCTICDIR; mkdir -p classes; cd src; javac -d ../classes *.java)

if [ "$2" = "lpc" ]; then
    echo Creating lpc files
    mkdir -p $VOICEDIR/lpc
    (cd $VOICEDIR
    bin/make_lpc wav/*.wav
    )

    echo Creating lpc/lpc.params
    for file in $VOICEDIR/lpc/*.lpc; do
	$ESTDIR/bin/ch_track -otype est_ascii $file
    done | sed '1,/EST_Header_End/d' |
    awk 'BEGIN {min=0; max=0;} {
         for (i=4; i<=NF; i++) {
             if ($i < min) min = $i;
             if ($i > max) max = $i;
         }
     } END {
         printf("LPC_MIN=%f\n",min);
         printf("LPC_MAX=%f\n",max);
         printf("LPC_RANGE=%f\n",max-min);
     }' > $VOICEDIR/lpc/lpc.params
fi

# build sts files
if [ "$2" = "sts" ]; then
 if [ "$FV_TYPE" = "diphone" ]; then
   # need to create scheme-formatted sts files for diphones
   echo "Finding STS files"
   . $VOICEDIR/lpc/lpc.params

   mkdir $VOICEDIR/sts 2>/dev/null

   # compile FindSTS
   CLASSFILES="FindSTS.class LPC.class STS.class Wave.class Utility.class"
   (cd $HELPERDIR
    javac FindSTS.java
    jar -cf FindSTS.jar $CLASSFILES
    rm -f $CLASSFILES 2>/dev/null)

   for f in $VOICEDIR/lpc/*.lpc; do
      fname=`basename $f .lpc`
      echo $fname STS
      java -cp "$HELPERDIR/FindSTS.jar" FindSTS $LPC_MIN $LPC_RANGE $f \
        $VOICEDIR/wav/$fname.wav $VOICEDIR/sts/$fname.sts
   done
 else
  # create STS files in the same format as for ARCTIC voices for clunits
  # and limited domain voices
    (cd $VOICEDIR
	echo "Creating short term signal (STS) files in sts/*.sts"
	mkdir -p sts
	java -cp $ARCTICDIR/classes FindSTS \
	    `find wav -type f | cut -f2 -d/ | cut -f1 -d.`
    )
 fi
fi

if [ "$2" = "mcep" ]; then
    # MCEP coefficients are not used for diphones
    echo Creating mcep/mcep.params and converting mcep files to text
    for file in $VOICEDIR/mcep/*.mcep; do
    echo $file MCEP
    $ESTDIR/bin/ch_track -otype est_ascii $file > $file.txt
    cat $file.txt
    done | sed '1,/EST_Header_End/d' |
    awk 'BEGIN {min=0; max=0;} {
         for (i=4; i<=NF; i++) {
             if ($i < min) min = $i;
             if ($i > max) max = $i;
         }
     } END {
         printf("MCEP_MIN=%f\n",min);
         printf("MCEP_MAX=%f\n",max);
         printf("MCEP_RANGE=%f\n",max-min);
     }' > $VOICEDIR/mcep/mcep.params
fi

idx_non_diphone() {
echo Creating unit index
(cd $VOICEDIR
    echo Creating $OUTDIR/misc.txt
    festival -b \
    festvox/$FV_FULLVOICENAME.scm \
	$ARCTICDIR/scheme/dump_misc.scm \
	"(begin (voice_${FV_FULLVOICENAME}) (dump_misc))" > \
	$OUTDIR/misc.txt

# UnitDatabase outputs its own info...
    java -cp $ARCTICDIR/classes UnitDatabase \
	festival/clunits/${FV_VOICENAME}.catalogue \
	`find wav -type f | cut -f2 -d/ | cut -f1 -d.`

echo Creating $OUTDIR/trees.txt
festival -b \
    festvox/$FV_FULLVOICENAME.scm \
    $ARCTICDIR/scheme/dump_trees.scm \
    "(begin (voice_${FV_FULLVOICENAME}) (dump_trees))" > \
    $OUTDIR/trees.txt

echo Creating $OUTDIR/weights.txt
festival -b \
    festvox/$FV_FULLVOICENAME.scm \
    $ARCTICDIR/scheme/dump_join_weights.scm \
    "(begin (voice_${FV_FULLVOICENAME}) (dump_join_weights))" > \
    $OUTDIR/weights.txt

echo Combining these files into $OUTDIR/$FV_VOICENAME.txt
(cd $OUTDIR; cat misc.txt unit_catalog.txt trees.txt unit_index.txt sts.txt mcep.txt weights.txt > $FV_VOICENAME.txt)

)
}

idx_diphone() {
    echo "Building diphone index"

    sed '1,/EST_Header_End/d' $VOICEDIR/dic/*.est |
    awk '{printf("( %s )\n",$0)}' >$VOICEDIR/dic/diphidx.unsorted.scm
    festival --heap 5000000 -b \
        $HELPERDIR/qsort.scm \
        '(begin (set! diphindex (load
            "'$VOICEDIR/dic/diphidx.unsorted.scm'" t))
         (set! diphindex (qsort diphindex carstring<? carstring=?))
         (while (not (null? diphindex))
            (set! x (car diphindex))
            (format t "( %l %l %l %l %l ) \n"
                (nth 0 x) (nth 1 x) (nth 2 x) (nth 3 x) (nth 4 x))
            (set! diphindex (cdr diphindex))
         ))' > $VOICEDIR/dic/diphidx.scm

    festival --heap 5000000 -b \
        $HELPERDIR/FestVoxDiphoneToFreeTTS.scm \
        '(dump_diphone "'$FV_VOICENAME'" "'$VOICEDIR'"
        "'$OUTDIR'" "header.txt" "data.txt"
        "'$VOICEDIR'/dic/diphidx.scm")'
        # parentheses allow script to only temporarily change to $OUTDIR
        (cd $OUTDIR
            rm -f README 2>/dev/null
            rm -f $FV_FULLVOICENAME.txt 2>/dev/null
            echo "The data for the voice $FV_FULLVOICENAME is stored in" > README
            echo "$FV_FULLVOICENAME.txt  All other files may be ignored." >> README
            echo "*** Generated by $0 $1 $2" > $FV_FULLVOICENAME.txt
            echo "*** clunits " `date` >> $FV_FULLVOICENAME.txt
            cat header.txt data.txt >> $FV_FULLVOICENAME.txt
        )

}

if [ "$2" = "idx" ]; then
    if [ "$FV_TYPE" = "diphone" ]; then
        idx_diphone
    else
        idx_non_diphone
    fi
fi

setGender() {
    while true; do
        echo "Please the number corresponding to the gender of this voice:"
        echo "     0 <Cancel>"
        echo "     1 Neutral"
        echo "     2 Male"
        echo "     3 Female"
        echo "     Q <Quit>: Abort the conversion process."
        read 

        if [ "$REPLY" = "0" ]; then
            break
        elif [ "$REPLY" = "1" ]; then
            VP_GENDER="NEUTRAL"