rockbox/tools/voice.pl

592 lines
19 KiB
Perl
Raw Normal View History

#!/usr/bin/perl -s
# __________ __ ___.
# Open \______ \ ____ ____ | | _\_ |__ _______ ___
# Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
# Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
# Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
# \/ \/ \/ \/ \/
# $Id$
#
# Copyright (C) 2007 Jonas Häggqvist
#
# All files in this archive are subject to the GNU General Public License.
# See the file COPYING in the source tree root for full license agreement.
#
# This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
# KIND, either express or implied.
use strict;
use warnings;
use File::Basename;
use File::Copy;
use vars qw($V $C $t $l $e $E $s $S $i $v $f);
use IPC::Open2;
use IPC::Open3;
use Digest::MD5 qw(md5_hex);
use DirHandle;
use open ':encoding(utf8)';
use open ':std';
use utf8;
sub printusage {
print <<USAGE
Usage: voice.pl [options] [path to dir]
-V
Create voice file. You must also specify -l, -i, and -t or -f
-C
Create .talk clips.
-t=<target>
Specify which target you want to build voicefile for. Must include
any features that target supports.
-f=<file> Use existing voiceids file
-i=<target_id>
Numeric target id. Needed for voice building.
-l=<language>
Specify which language you want to build. Without .lang extension.
-e=<encoder>
Which encoder to use for voice strings
-E=<encoder options>
Which encoder options to use when compressing voice strings. Enclose
in double quotes if the options include spaces.
-s=<TTS engine>
Which TTS engine to use.
-S=<TTS engine options>
Options to pass to the TTS engine. Enclose in double quotes if the
options include spaces.
-v
Be verbose
USAGE
;
}
my %festival_lang_map = (
'english' => 'english',
'english-us' => 'english',
'espanol' => 'spanish',
#'finnish' => 'finnish'
#'italiano' => 'italian',
#'czech' => 'czech',
#'welsh' => 'welsh'
);
my %gtts_lang_map = (
'english' => 'en-gb', # Always first, it's the golden master
'czech' => 'cs', # not supported
'deutsch' => 'de',
'english-us' => 'en-us',
'espanol' => 'es-es',
'francais' => 'fr-fr',
'greek' => 'el',
'magyar' => 'hu',
'italiano' => 'it',
'nederlands' => 'nl',
'norsk' => 'no',
'polski' => 'pl',
'russian' => 'ru',
'slovak' => 'sk',
'srpski' => 'sr',
'svenska' => 'sv',
'turkce' => 'tr',
);
my %espeak_lang_map = (
'english' => 'en-gb', # Always first, it's the golden master
'czech' => 'cs',
'deutsch' => 'de',
'english-us' => 'en-us',
'espanol' => 'es',
'francais' => 'fr-fr',
'greek' => 'el',
'nederlands' => 'nl',
'magyar' => 'hu',
'italiano' => 'it',
'japanese' => 'ja',
'nederlands' => 'nl',
'norsk' => 'no',
'polski' => 'pl',
'russian' => 'ru',
'slovak' => 'sk',
'srpski' => 'sr',
'svenska' => 'sv',
'turkce' => 'tr',
);
# Initialize TTS engine. May return an object or value which will be passed
# to voicestring and shutdown_tts
sub init_tts {
our $verbose;
my ($tts_engine, $tts_engine_opts, $language) = @_;
my %ret = ("name" => $tts_engine);
$ret{"format"} = 'wav';
$ret{"ttsoptions"} = "";
# Don't use given/when here - it's not compatible with old perl versions
if ($tts_engine eq 'festival') {
print("> festival $tts_engine_opts --server\n") if $verbose;
my $pid = open(FESTIVAL_SERVER, "| festival $tts_engine_opts --server > /dev/null 2>&1");
my $dummy = *FESTIVAL_SERVER; #suppress warning
$SIG{INT} = sub { kill TERM => $pid; print("foo"); panic_cleanup(); };
$SIG{KILL} = sub { kill TERM => $pid; print("boo"); panic_cleanup(); };
$ret{"pid"} = $pid;
if (defined($festival_lang_map{$language}) && $tts_engine_opts !~ /--language/) {
$ret{"ttsoptions"} = "--language $festival_lang_map{$language} ";
}
} elsif ($tts_engine eq 'sapi') {
my $toolsdir = dirname($0);
my $path = `cygpath $toolsdir -a -w`;
chomp($path);
$path = $path . '\\';
my $cmd = $path . "sapi_voice.vbs /language:$language $tts_engine_opts";
$cmd =~ s/\\/\\\\/g;
print("> cscript //nologo $cmd\n") if $verbose;
my $pid = open2(*CMD_OUT, *CMD_IN, "cscript //nologo $cmd");
binmode(*CMD_IN, ':encoding(utf16le)');
binmode(*CMD_OUT, ':encoding(utf16le)');
$SIG{INT} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
$SIG{KILL} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
print(CMD_IN "QUERY\tVENDOR\r\n");
my $vendor = readline(*CMD_OUT);
$vendor =~ s/\r\n//;
%ret = (%ret,
"stdin" => *CMD_IN,
"stdout" => *CMD_OUT,
"vendor" => $vendor);
} elsif ($tts_engine eq 'gtts') {
$ret{"format"} = 'mp3';
if (defined($gtts_lang_map{$language}) && $tts_engine_opts !~ /-l/) {
$ret{"ttsoptions"} = "-l $gtts_lang_map{$language} ";
}
} elsif ($tts_engine eq 'espeak' || $tts_engine eq 'espeak-ng') {
if (defined($espeak_lang_map{$language}) && $tts_engine_opts !~ /-v/) {
$ret{"ttsoptions"} = "-v$espeak_lang_map{$language} ";
}
}
return \%ret;
}
# Shutdown TTS engine if necessary.
sub shutdown_tts {
my ($tts_object) = @_;
if ($$tts_object{'name'} eq 'festival') {
# Send SIGTERM to festival server
kill TERM => $$tts_object{"pid"};
}
elsif ($$tts_object{'name'} eq 'sapi') {
print({$$tts_object{"stdin"}} "QUIT\r\n");
close($$tts_object{"stdin"});
}
}
# Apply corrections to a voice-string to make it sound better
sub correct_string {
our $verbose;
my ($string, $language, $tts_object) = @_;
my $orig = $string;
my $corrections = $tts_object->{"corrections"};
foreach (@$corrections) {
my $r = "s" . $_->{separator} . $_->{search} . $_->{separator}
. $_->{replace} . $_->{separator} . $_->{modifier};
eval ('$string =~' . "$r;");
}
if ($orig ne $string) {
printf("%s -> %s\n", $orig, $string) if $verbose;
}
return $string;
}
# Produce a wav file of the text given
sub voicestring {
our $verbose;
my ($string, $output, $tts_engine_opts, $tts_object) = @_;
my $cmd;
my $name = $$tts_object{'name'};
$tts_engine_opts .= $$tts_object{"ttsoptions"};
printf("Generate \"%s\" with %s in file %s\n", $string, $name, $output) if $verbose;
if ($name eq 'festival') {
# festival_client lies to us, so we have to do awful soul-eating
# work with IPC::open3()
$cmd = "festival_client --server localhost --otype riff --ttw --output \"$output\"";
# Use festival-prolog.scm if it's there (created by user of tools/configure)
if (-f "festival-prolog.scm") {
$cmd .= " --prolog festival-prolog.scm";
}
print("> $cmd\n") if $verbose;
# Open command, and filehandles for STDIN, STDOUT, STDERR
my $pid = open3(*CMD_IN, *CMD_OUT, *CMD_ERR, $cmd);
# Put the string to speak into STDIN and close it
print(CMD_IN $string);
close(CMD_IN);
# Read all output from festival_client (because it LIES TO US)
while (<CMD_ERR>) {
}
close(CMD_OUT);
close(CMD_ERR);
}
elsif ($name eq 'flite') {
$cmd = "flite $tts_engine_opts -t \"$string\" \"$output\"";
print("> $cmd\n") if $verbose;
system($cmd);
}
elsif ($name eq 'espeak') {
$cmd = "espeak $tts_engine_opts -w \"$output\" --stdin";
print("> $cmd\n") if $verbose;
open(RBSPEAK, "| $cmd");
print RBSPEAK $string . "\n";
close(RBSPEAK);
}
elsif ($name eq 'espeak-ng') {
$cmd = "espeak-ng $tts_engine_opts -w \"$output\" --stdin";
print("> $cmd\n") if $verbose;
open(RBSPEAK, "| $cmd");
print RBSPEAK $string . "\n";
close(RBSPEAK);
}
elsif ($name eq 'sapi') {
print({$$tts_object{"stdin"}} "SPEAK\t$output\t$string\r\n");
}
elsif ($name eq 'swift') {
$cmd = "swift $tts_engine_opts -o \"$output\" \"$string\"";
print("> $cmd\n") if $verbose;
system($cmd);
}
elsif ($name eq 'rbspeak') {
# xxx: $tts_engine_opts isn't used
$cmd = "rbspeak $output";
print("> $cmd\n") if $verbose;
open(RBSPEAK, "| $cmd");
print RBSPEAK $string . "\n";
close(RBSPEAK);
}
elsif ($name eq 'mimic') {
$cmd = "mimic $tts_engine_opts -o $output";
print("> $cmd\n") if $verbose;
open(RBSPEAK, "| $cmd");
print RBSPEAK $string . "\n";
close(RBSPEAK);
}
elsif ($name eq 'gtts') {
$cmd = "gtts-cli $tts_engine_opts -o $output -";
print("> $cmd\n") if $verbose;
open(RBSPEAK, "| $cmd");
print RBSPEAK $string . "\n";
close(RBSPEAK);
}
}
# trim leading / trailing silence from the clip
sub wavtrim {
our $verbose;
my ($file, $threshold, $tts_object) = @_;
printf("Trim \"%s\"\n", $file) if $verbose;
my $cmd = "wavtrim \"$file\" $threshold";
if ($$tts_object{"name"} eq "sapi") {
print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
}
else {
print("> $cmd\n") if $verbose;
`$cmd`;
}
}
# Encode a wav file into the given destination file
sub encodewav {
our $verbose;
my ($input, $output, $encoder, $encoder_opts, $tts_object) = @_;
printf("Encode \"%s\" with %s in file %s\n", $input, $encoder, $output) if $verbose;
my $cmd = "$encoder $encoder_opts \"$input\" \"$output\"";
if ($$tts_object{"name"} eq "sapi") {
print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
}
else {
print("> $cmd\n") if $verbose;
`$cmd`;
}
}
# synchronize the clip generation / processing if it's running in another process
sub synchronize {
my ($tts_object) = @_;
if ($$tts_object{"name"} eq "sapi") {
print({$$tts_object{"stdin"}} "SYNC\t42\r\n");
my $wait = readline($$tts_object{"stdout"});
#ignore what's actually returned
}
}
# Run genlang and create voice clips for each string
sub generateclips {
our $verbose;
my ($language, $target, $encoder, $encoder_opts, $tts_engine, $tts_engine_opts, $existingids) = @_;
my $english = dirname($0) . '/../apps/lang/english.lang';
my $langfile = dirname($0) . '/../apps/lang/' . $language . '.lang';
my $correctionsfile = dirname($0) . '/voice-corrections.txt';
my $idfile = "$language.vid";
my $updfile = "$language-update.lang";
my $id = '';
my $voice = '';
my $cmd;
my $pool_file;
my $i = 0;
local $| = 1; # make progress indicator work reliably
# First run the language through an update pass so any missing strings
# are backfilled from English. Without this, BADNESS.
if ($existingids) {
$idfile = $existingids;
} else {
$cmd = "updatelang $english $langfile $updfile";
print("> $cmd\n") if $verbose;
system($cmd);
$cmd = "genlang -o -t=$target -e=$english $updfile 2>/dev/null > $idfile";
print("> $cmd\n") if $verbose;
system($cmd);
}
open(VOICEFONTIDS, " < $idfile");
my $tts_object = init_tts($tts_engine, $tts_engine_opts, $language);
# add string corrections to tts_object.
my @corrects = ();
open(VOICEREGEXP, "<$correctionsfile") or die "Can't open corrections file!\n";
while(<VOICEREGEXP>) {
# get first character of line
my $line = $_;
my $separator = substr($_, 0, 1);
if($separator =~ m/\s+/) {
next;
}
chomp($line);
$line =~ s/^.//g; # remove separator at beginning
my ($lang, $engine, $vendor, $search, $replace, $modifier) = split(/$separator/, $line);
# does language match?
if($language !~ m/$lang/) {
next;
}
if($$tts_object{"name"} !~ m/$engine/) {
next;
}
my $v = $$tts_object{"vendor"} || ""; # vendor might be empty in $tts_object
if($v !~ m/$vendor/) {
next;
}
push @corrects, {separator => $separator, search => $search, replace => $replace, modifier => $modifier};
}
close(VOICEREGEXP);
$tts_object->{corrections} = [@corrects];
print("Generating voice clips");
print("\n") if $verbose;
for (<VOICEFONTIDS>) {
my $line = $_;
if ($line =~ /^id: (.*)$/) {
$id = $1;
}
elsif ($line =~ /^voice: "(.*)"$/) {
$voice = $1;
if ($id !~ /^NOT_USED_.*$/ && $voice ne "") {
my $wav = $id . '.wav';
my $enc = $id . '.mp3';
# Print some progress information
if (++$i % 10 == 0 and !$verbose) {
print(".");
}
# Apply corrections to the string
$voice = correct_string($voice, $language, $tts_object);
# If we have a pool of snippets, see if the string exists there first
if (defined($ENV{'POOL'})) {
$pool_file = sprintf("%s/%s-%s.mp3", $ENV{'POOL'},
md5_hex(Encode::encode_utf8("$voice $tts_engine $tts_engine_opts $encoder_opts")),
$language);
if (-f $pool_file) {
printf("Re-using %s (%s) from pool\n", $id, $voice) if $verbose;
copy($pool_file, $enc);
}
}
# Don't generate encoded file if it already exists (probably from the POOL)
if (! -f $enc) {
if ($id eq "VOICE_PAUSE") {
print("Use distributed $wav\n") if $verbose;
copy(dirname($0)."/VOICE_PAUSE.wav", $wav);
} else {
voicestring($voice, $wav, $tts_engine_opts, $tts_object);
if ($tts_object->{'format'} eq "wav") {
wavtrim($wav, 500, $tts_object);
# 500 seems to be a reasonable default for now
}
}
if ($tts_object->{'format'} eq "wav" || $id eq "VOICE_PAUSE") {
encodewav($wav, $enc, $encoder, $encoder_opts, $tts_object);
} else {
copy($wav, $enc);
}
synchronize($tts_object);
if (defined($ENV{'POOL'})) {
copy($enc, $pool_file);
}
unlink($wav);
}
$voice = "";
$id = "";
}
}
}
close(VOICEFONTIDS);
print("\n");
unlink($updfile) if (-f $updfile);
shutdown_tts($tts_object);
}
# Assemble the voicefile
sub createvoice {
our $verbose;
my ($language, $target_id, $existingids) = @_;
my $outfile = "";
my $vfile = "$language.vid";
if ($existingids) {
$vfile = $existingids;
}
$outfile = sprintf("%s.voice", $language);
printf("Saving voice file to %s\n", $outfile) if $verbose;
my $cmd = "voicefont '$vfile' $target_id ./ $outfile";
print("> $cmd\n") if $verbose;
my $output = `$cmd`;
print($output) if $verbose;
if (!$existingids) {
unlink("$vfile");
}
}
sub deleteencs() {
for (glob('*.mp3')) {
unlink($_);
}
for (glob('*.wav')) {
unlink($_);
}
}
sub panic_cleanup {
deletencs();
die "moo";
}
# Generate .talk clips
sub gentalkclips {
our $verbose;
my ($dir, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i) = @_;
my $d = new DirHandle $dir;
while (my $file = $d->read) {
my ($voice, $wav, $enc);
# Print some progress information
if (++$i % 10 == 0 and !$verbose) {
print(".");
}
# Convert to a complete path
my $path = sprintf("%s/%s", $dir, $file);
$voice = $file;
$wav = sprintf("%s.talk.wav", $path);
# Ignore dot-dirs and talk files
if ($file eq '.' || $file eq '..' || $file =~ /\.talk$/) {
next;
}
# Element is a dir
if ( -d $path) {
gentalkclips($path, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i);
$enc = sprintf("%s/_dirname.talk", $path);
}
# Element is a file
else {
$enc = sprintf("%s.talk", $path);
$voice =~ s/\.[^\.]*$//; # Trim extension
}
printf("Talkclip %s: %s", $enc, $voice) if $verbose;
voicestring($voice, $wav, $tts_engine_opts, $tts_object);
wavtrim($wav, 500, $tts_object);
# 500 seems to be a reasonable default for now
encodewav($wav, $enc, $encoder, $encoder_opts, $tts_object);
synchronize($tts_object);
unlink($wav);
}
}
# Check parameters
my $printusage = 0;
unless (defined($V) or defined($C)) { print("Missing either -V or -C\n"); $printusage = 1; }
if (defined($V)) {
unless (defined($l)) { print("Missing -l argument\n"); $printusage = 1; }
unless (defined($i)) { print("Missing -i argument\n"); $printusage = 1; }
if (defined($t) && defined($f) ||
!defined($t) && !defined($f)) {
print("Missing either -t or -f argument\n"); $printusage = 1;
}
}
elsif (defined($C)) {
unless (defined($ARGV[0])) { print "Missing path argument\n"; $printusage = 1; }
}
unless (defined($e)) { print("Missing -e argument\n"); $printusage = 1; }
unless (defined($E)) { print("Missing -E argument\n"); $printusage = 1; }
unless (defined($s)) { print("Missing -s argument\n"); $printusage = 1; }
unless (defined($S)) { print("Missing -S argument\n"); $printusage = 1; }
if ($printusage == 1) { printusage(); exit 1; }
if (defined($v) or defined($ENV{'V'})) {
our $verbose = 1;
}
# add the tools dir to the path temporarily, for calling various tools
$ENV{'PATH'} = dirname($0) . ':' . $ENV{'PATH'};
# Do what we're told
if ($V == 1) {
# Only do the panic cleanup for voicefiles
$SIG{INT} = \&panic_cleanup;
$SIG{KILL} = \&panic_cleanup;
printf("Generating voice\n Target: %s\n Language: %s\n Encoder (options): %s (%s)\n TTS Engine (options): %s (%s)\n",
defined($t) ? $t : "unknown",
$l, $e, $E, $s, $S);
generateclips($l, $t, $e, $E, $s, $S, $f);
createvoice($l, $i, $f);
deleteencs();
}
elsif ($C) {
printf("Generating .talk clips\n Path: %s\n Language: %s\n Encoder (options): %s (%s)\n TTS Engine (options): %s (%s)\n", $ARGV[0], $l, $e, $E, $s, $S);
my $tts_object = init_tts($s, $S, $l);
gentalkclips($ARGV[0], $tts_object, $e, $E, $S, 0);
shutdown_tts($tts_object);
}
else {
printusage();
exit 1;
}