tools/voice.pl

   1 #!/usr/bin/perl -s
   2 #             __________               __   ___.
   3 #   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
   4 #   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
   5 #   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
   6 #   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
   7 #                     \/            \/     \/    \/            \/
   8 # $Id$
   9 #
  10 # Copyright (C) 2007 Jonas Häggqvist
  11 #
  12 # All files in this archive are subject to the GNU General Public License.
  13 # See the file COPYING in the source tree root for full license agreement.
  14 #
  15 # This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
  16 # KIND, either express or implied.
  17
  18 use strict;
  19 use warnings;
  20 use File::Basename;
  21 use File::Copy;
  22 use Switch;
  23 use vars qw($V $C $t $l $e $E $s $S $i $v);
  24 use IPC::Open2;
  25 use IPC::Open3;
  26 use Digest::MD5 qw(md5_hex);
  27 use DirHandle;
  28 use open ':encoding(utf8)';
  29 use open ':std';
  30
  31 sub printusage {
  32     print <<USAGE
  33
  34 Usage: voice.pl [options] [path to dir]
  35  -V
  36     Create voice file. You must also specify -t and -l.
  37
  38  -C
  39     Create .talk clips.
  40
  41  -t=<target>
  42     Specify which target you want to build voicefile for. Must include
  43     any features that target supports.
  44
  45  -i=<target_id>
  46     Numeric target id. Needed for voice building.
  47
  48  -l=<language>
  49     Specify which language you want to build. Without .lang extension.
  50
  51  -e=<encoder>
  52     Which encoder to use for voice strings
  53
  54  -E=<encoder options>
  55     Which encoder options to use when compressing voice strings. Enclose
  56     in double quotes if the options include spaces.
  57
  58  -s=<TTS engine>
  59     Which TTS engine to use.
  60
  61  -S=<TTS engine options>
  62     Options to pass to the TTS engine. Enclose in double quotes if the
  63     options include spaces.
  64
  65  -v
  66     Be verbose
  67 USAGE
  68 ;
  69 }
  70
  71 # Initialize TTS engine. May return an object or value which will be passed
  72 # to voicestring and shutdown_tts
  73 sub init_tts {
  74     our $verbose;
  75     my ($tts_engine, $tts_engine_opts, $language) = @_;
  76     my %ret = ("name" => $tts_engine);
  77     switch($tts_engine) {
  78         case "festival" {
  79             print("> festival $tts_engine_opts --server\n") if $verbose;
  80             my $pid = open(FESTIVAL_SERVER, "| festival $tts_engine_opts --server > /dev/null 2>&1");
  81             my $dummy = *FESTIVAL_SERVER; #suppress warning
  82             $SIG{INT} = sub { kill TERM => $pid; print("foo"); panic_cleanup(); };
  83             $SIG{KILL} = sub { kill TERM => $pid; print("boo"); panic_cleanup(); };
  84             $ret{"pid"} = $pid;
  85         }
  86         case "sapi" {
  87             my $toolsdir = dirname($0);
  88             my $path = `cygpath $toolsdir -a -w`;
  89             chomp($path);
  90             $path = $path . '\\';
  91             my $cmd = $path . "sapi_voice.vbs /language:$language $tts_engine_opts";
  92             $cmd =~ s/\\/\\\\/g;
  93             print("> cscript //nologo $cmd\n") if $verbose;
  94             my $pid = open2(*CMD_OUT, *CMD_IN, "cscript //nologo $cmd");
  95             binmode(*CMD_IN, ':encoding(utf16le)');
  96             binmode(*CMD_OUT, ':encoding(utf16le)');
  97             $SIG{INT} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
  98             $SIG{KILL} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
  99             print(CMD_IN "QUERY\tVENDOR\r\n");
 100             my $vendor = readline(*CMD_OUT);
 101             $vendor =~ s/\r\n//;
 102             %ret = (%ret,
 103                     "stdin" => *CMD_IN,
 104                     "stdout" => *CMD_OUT,
 105                     "vendor" => $vendor);
 106         }
 107     }
 108     return \%ret;
 109 }
 110
 111 # Shutdown TTS engine if necessary.
 112 sub shutdown_tts {
 113     my ($tts_object) = @_;
 114     switch($$tts_object{"name"}) {
 115         case "festival" {
 116             # Send SIGTERM to festival server
 117             kill TERM => $$tts_object{"pid"};
 118         }
 119         case "sapi" {
 120             print({$$tts_object{"stdin"}} "QUIT\r\n");
 121             close($$tts_object{"stdin"});
 122         }
 123     }
 124 }
 125
 126 # Apply corrections to a voice-string to make it sound better
 127 sub correct_string {
 128     our $verbose;
 129     my ($string, $language, $tts_object) = @_;
 130     my $orig = $string;
 131     switch($language) {
 132         # General for all engines and languages
 133         $string =~ s/USB/U S B/g;
 134         $string =~ s/ID3/I D 3/g;
 135
 136         case "english" {
 137             switch($$tts_object{"name"}) {
 138                 case ["sapi","festival"] {
 139                     $string =~ s/plugin(s?)/plug-in$1/ig; next
 140                 }
 141                 case "festival" {
 142                     $string =~ s/\ba\b/ay/ig;
 143                 }
 144             }
 145         }
 146         case "deutsch" {
 147             # for all german engines (e.g. for english words)
 148             $string =~ s/alkaline/alkalein/ig;
 149             $string =~ s/byte(s?)/beit$1/ig;
 150             $string =~ s/clip(s?)/klipp$1/ig;
 151             $string =~ s/cuesheet/kjuschiet/ig;
 152             $string =~ s/dither/didder/ig;
 153             $string =~ s/equalizer/iquileiser/ig;
 154             $string =~ s/\bflash\b/fläsh/ig;
 155             $string =~ s/\bfirmware(s?)\b/firmwer$1/ig;
 156             $string =~ s/\bI D 3 tag\b/I D 3 täg/ig; # can't just use "tag" here
 157             $string =~ s/\bloudness\b/laudness/ig;
 158             $string =~ s/\bunicode\b/unikod/ig;
 159             switch($$tts_object{"name"}) {
 160                  case "sapi" {   # just for SAPI
 161                     switch($$tts_object{"vendor"}) {
 162                         case "AT&T Labs" {
 163                             $string =~ s/alphabet/alfabet/ig;
 164                             $string =~ s/ampere/amper/ig;
 165                             $string =~ s/\bdezibel\b/de-zibell/ig;
 166                             $string =~ s/diddering/didde-ring/ig;
 167                             $string =~ s/energie\b/ener-gie/ig;
 168                             $string =~ s/\Blauf\b/-lauf/ig;
 169                             $string =~ s/\bnumerisch\b/numehrisch/ig;
 170                         }
 171                     }
 172                 }
 173             }
 174         }
 175         case "svenska" {
 176             # for all swedish engines (e.g. for english words)
 177             $string =~ s/kilobyte/kilobajt/ig;
 178             $string =~ s/megabyte/megabajt/ig;
 179             $string =~ s/gigabyte/gigabajt/ig;
 180             $string =~ s/\bloudness\b/laudness/ig;
 181
 182             switch($$tts_object{"name"}) {
 183                  case "espeak" {   # just for eSpeak
 184                      $string =~ s/ampere/ampär/ig;
 185                      $string =~ s/bokmärken/bok-märken/ig;
 186                      $string =~ s/generella/schenerella/ig;
 187                      $string =~ s/dithering/diddering/ig;
 188                      $string =~ s/\bunicode\b/jynikod/ig;
 189                      $string =~ s/uttoning/utoning/ig;
 190                      $string =~ s/procent/pro-cent/ig;
 191                      $string =~ s/spellistor/spelistor/ig;
 192                      $string =~ s/cuesheet/qjyschiit/ig;
 193                  }
 194             }
 195         }
 196         case "italiano" {
 197             # for all italian engines (e.g. for english words)
 198             $string =~ s/Replaygain/Ripleyghein/ig;
 199             $string =~ s/Crossfade/Crossfeid/ig;
 200             $string =~ s/beep/Bip/ig;
 201             $string =~ s/cuesheet/chiushit/ig;
 202             $string =~ s/fade/feid/ig;
 203             $string =~ s/Crossfeed/crossfid/ig;
 204             $string =~ s/Cache/chash/ig;
 205             $string =~ s/\bfirmware(s?)\b/firmuer$1/ig;
 206             $string =~ s/\bFile(s?)\b/fail$1/ig;
 207             $string =~ s/\bloudness\b/laudness/ig;
 208             $string =~ s/\bunicode\b/unikod/ig;
 209             $string =~ s/Playlist/pleylist/ig;
 210             $string =~ s/WavPack/wave pak/ig;
 211             $string =~ s/BITRATE/bit reit/ig;
 212             $string =~ s/Codepage/cod page/ig;
 213             $string =~ s/PCM Wave/pcm Ue'iv/ig;
 214             switch($$tts_object{"name"}) {
 215                  case "sapi" {   # just for SAPI
 216                     switch($$tts_object{"vendor"}) {
 217                         case "Loquendo" {
 218                             $string =~ s/Inizializza/inizializa/ig;
 219                         }
 220                         case "ScanSoft, Inc" {
 221                             $string =~ s/V/v/ig;
 222                             $string =~ s/X/x/ig;
 223                             $string =~ s/stop/stohp/ig;
 224                         }
 225                     }
 226                 }
 227             }
 228         }
 229     }
 230     if ($orig ne $string) {
 231         printf("%s -> %s\n", $orig, $string) if $verbose;
 232     }
 233     return $string;
 234 }
 235
 236 # Produce a wav file of the text given
 237 sub voicestring {
 238     our $verbose;
 239     my ($string, $output, $tts_engine_opts, $tts_object) = @_;
 240     my $cmd;
 241     printf("Generate \"%s\" with %s in file %s\n", $string, $$tts_object{"name"}, $output) if $verbose;
 242     switch($$tts_object{"name"}) {
 243         case "festival" {
 244             # festival_client lies to us, so we have to do awful soul-eating
 245             # work with IPC::open3()
 246             $cmd = "festival_client --server localhost --otype riff --ttw --output \"$output\"";
 247             print("> $cmd\n") if $verbose;
 248             # Open command, and filehandles for STDIN, STDOUT, STDERR
 249             my $pid = open3(*CMD_IN, *CMD_OUT, *CMD_ERR, $cmd);
 250             # Put the string to speak into STDIN and close it
 251             print(CMD_IN $string);
 252             close(CMD_IN);
 253             # Read all output from festival_client (because it LIES TO US)
 254             while (<CMD_ERR>) {
 255             }
 256             close(CMD_OUT);
 257             close(CMD_ERR);
 258         }
 259         case "flite" {
 260             $cmd = "flite $tts_engine_opts -t \"$string\" \"$output\"";
 261             print("> $cmd\n") if $verbose;
 262             `$cmd`;
 263         }
 264         case "espeak" {
 265             $cmd = "espeak $tts_engine_opts -w \"$output\"";
 266             print("> $cmd\n") if $verbose;
 267             open(ESPEAK, "| $cmd");
 268             print ESPEAK $string . "\n";
 269             close(ESPEAK);
 270         }
 271         case "sapi" {
 272             print({$$tts_object{"stdin"}} "SPEAK\t$output\t$string\r\n");
 273         }
 274         case "swift" {
 275             $cmd = "swift $tts_engine_opts -o \"$output\" \"$string\"";
 276             print("> $cmd\n") if $verbose;
 277             system($cmd);
 278         }
 279     }
 280 }
 281
 282 # trim leading / trailing silence from the clip
 283 sub wavtrim {
 284     our $verbose;
 285     my ($file, $threshold, $tts_object) = @_;
 286     printf("Trim \"%s\"\n", $file) if $verbose;
 287     my $cmd = "wavtrim \"$file\" $threshold";
 288     if ($$tts_object{"name"} eq "sapi") {
 289         print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
 290     }
 291     else {
 292         print("> $cmd\n") if $verbose;
 293         `$cmd`;
 294     }
 295 }
 296
 297 # Encode a wav file into the given destination file
 298 sub encodewav {
 299     our $verbose;
 300     my ($input, $output, $encoder, $encoder_opts, $tts_object) = @_;
 301     printf("Encode \"%s\" with %s in file %s\n", $input, $encoder, $output) if $verbose;
 302     my $cmd = "$encoder $encoder_opts \"$input\" \"$output\"";
 303     if ($$tts_object{"name"} eq "sapi") {
 304         print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
 305     }
 306     else {
 307         print("> $cmd\n") if $verbose;
 308         `$cmd`;
 309     }
 310 }
 311
 312 # synchronize the clip generation / processing if it's running in another process
 313 sub synchronize {
 314     my ($tts_object) = @_;
 315     if ($$tts_object{"name"} eq "sapi") {
 316         print({$$tts_object{"stdin"}} "SYNC\t42\r\n");
 317         my $wait = readline($$tts_object{"stdout"});
 318         #ignore what's actually returned
 319     }
 320 }
 321
 322 # Run genlang and create voice clips for each string
 323 sub generateclips {
 324     our $verbose;
 325     my ($language, $target, $encoder, $encoder_opts, $tts_engine, $tts_engine_opts) = @_;
 326     my $english = dirname($0) . '/../apps/lang/english.lang';
 327     my $langfile = dirname($0) . '/../apps/lang/' . $language . '.lang';
 328     my $id = '';
 329     my $voice = '';
 330     my $cmd = "genlang -o -t=$target -e=$english $langfile 2>/dev/null";
 331     my $pool_file;
 332     open(VOICEFONTIDS, "> voicefontids");
 333     my $i = 0;
 334     local $| = 1; # make progress indicator work reliably
 335
 336     my $tts_object = init_tts($tts_engine, $tts_engine_opts, $language);
 337     print("Generating voice clips");
 338     print("\n") if $verbose;
 339     for (`$cmd`) {
 340         my $line = $_;
 341         print(VOICEFONTIDS $line);
 342         if ($line =~ /^id: (.*)$/) {
 343             $id = $1;
 344         }
 345         elsif ($line =~ /^voice: "(.*)"$/) {
 346             $voice = $1;
 347             if ($id !~ /^NOT_USED_.*$/ && $voice ne "") {
 348                 my $wav = $id . '.wav';
 349                 my $mp3 = $id . '.mp3';
 350
 351                 # Print some progress information
 352                 if (++$i % 10 == 0 and !$verbose) {
 353                     print(".");
 354                 }
 355
 356                 # Apply corrections to the string
 357                 $voice = correct_string($voice, $language, $tts_object);
 358
 359                 # If we have a pool of snippets, see if the string exists there first
 360                 if (defined($ENV{'POOL'})) {
 361                     $pool_file = sprintf("%s/%s-%s.mp3", $ENV{'POOL'},
 362                                          md5_hex("$voice $tts_engine $tts_engine_opts $encoder_opts"),
 363                                          $language);
 364                     if (-f $pool_file) {
 365                         printf("Re-using %s (%s) from pool\n", $id, $voice) if $verbose;
 366                         copy($pool_file, $mp3);
 367                     }
 368                 }
 369
 370                 # Don't generate MP3 if it already exists (probably from the POOL)
 371                 if (! -f $mp3) {
 372                     if ($id eq "VOICE_PAUSE") {
 373                         print("Use distributed $wav\n") if $verbose;
 374                         copy(dirname($0)."/VOICE_PAUSE.wav", $wav);
 375                     }
 376                     else {
 377                         voicestring($voice, $wav, $tts_engine_opts, $tts_object);
 378                         wavtrim($wav, 500, $tts_object);
 379                         # 500 seems to be a reasonable default for now
 380                     }
 381
 382                     encodewav($wav, $mp3, $encoder, $encoder_opts, $tts_object);
 383                     synchronize($tts_object);
 384                     if (defined($ENV{'POOL'})) {
 385                         copy($mp3, $pool_file);
 386                     }
 387                     unlink($wav);
 388                 }
 389                 $voice = "";
 390                 $id = "";
 391             }
 392         }
 393     }
 394     print("\n");
 395     close(VOICEFONTIDS);
 396     shutdown_tts($tts_object);
 397 }
 398
 399 # Assemble the voicefile
 400 sub createvoice {
 401     our $verbose;
 402     my ($language, $target_id) = @_;
 403     my $outfile = "";
 404     $outfile = sprintf("%s.voice", $language);
 405     printf("Saving voice file to %s\n", $outfile) if $verbose;
 406     my $cmd = "voicefont 'voicefontids' $target_id ./ $outfile";
 407     print("> $cmd\n") if $verbose;
 408     my $output = `$cmd`;
 409     print($output) if $verbose;
 410 }
 411
 412 sub deletemp3s() {
 413     for (glob('*.mp3')) {
 414         unlink($_);
 415     }
 416     for (glob('*.wav')) {
 417         unlink($_);
 418     }
 419 }
 420
 421 sub panic_cleanup {
 422     deletemp3s();
 423     die "moo";
 424 }
 425
 426 # Generate .talk clips
 427 sub gentalkclips {
 428     our $verbose;
 429     my ($dir, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i) = @_;
 430     my $d = new DirHandle $dir;
 431     while (my $file = $d->read) {
 432         my ($voice, $wav, $mp3);
 433         # Print some progress information
 434         if (++$i % 10 == 0 and !$verbose) {
 435             print(".");
 436         }
 437
 438         # Convert to a complete path
 439         my $path = sprintf("%s/%s", $dir, $file);
 440
 441         $voice = $file;
 442         $wav = sprintf("%s.talk.wav", $path);
 443
 444         # Ignore dot-dirs and talk files
 445         if ($file eq '.' || $file eq '..' || $file =~ /\.talk$/) {
 446             next;
 447         }
 448         # Element is a dir
 449         if ( -d $path) {
 450             gentalkclips($path, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i);
 451             $mp3 = sprintf("%s/_dirname.talk", $path);
 452         }
 453         # Element is a file
 454         else {
 455             $mp3 = sprintf("%s.talk", $path);
 456             $voice =~ s/\.[^\.]*$//; # Trim extension
 457         }
 458
 459         printf("Talkclip %s: %s", $mp3, $voice) if $verbose;
 460
 461         voicestring($voice, $wav, $tts_engine_opts, $tts_object);
 462         wavtrim($wav, 500, $tts_object);
 463         # 500 seems to be a reasonable default for now
 464         encodewav($wav, $mp3, $encoder, $encoder_opts, $tts_object);
 465         synchronize($tts_object);
 466         unlink($wav);
 467     }
 468 }
 469
 470
 471 # Check parameters
 472 my $printusage = 0;
 473 unless (defined($V) or defined($C)) { print("Missing either -V or -C\n"); $printusage = 1; }
 474 if (defined($V)) {
 475     unless (defined($t)) { print("Missing -t argument\n"); $printusage = 1; }
 476     unless (defined($l)) { print("Missing -l argument\n"); $printusage = 1; }
 477     unless (defined($i)) { print("Missing -i argument\n"); $printusage = 1; }
 478 }
 479 elsif (defined($C)) {
 480     unless (defined($ARGV[0])) { print "Missing path argument\n"; $printusage = 1; }
 481 }
 482 unless (defined($e)) { print("Missing -e argument\n"); $printusage = 1; }
 483 unless (defined($E)) { print("Missing -E argument\n"); $printusage = 1; }
 484 unless (defined($s)) { print("Missing -s argument\n"); $printusage = 1; }
 485 unless (defined($S)) { print("Missing -S argument\n"); $printusage = 1; }
 486 if ($printusage == 1) { printusage(); exit 1; }
 487
 488 if (defined($v) or defined($ENV{'V'})) {
 489     our $verbose = 1;
 490 }
 491
 492 # add the tools dir to the path temporarily, for calling various tools
 493 $ENV{'PATH'} = dirname($0) . ':' . $ENV{'PATH'};
 494
 495
 496 # Do what we're told
 497 if ($V == 1) {
 498     # Only do the panic cleanup for voicefiles
 499     $SIG{INT} = \&panic_cleanup;
 500     $SIG{KILL} = \&panic_cleanup;
 501
 502     printf("Generating voice\n  Target: %s\n  Language: %s\n  Encoder (options): %s (%s)\n  TTS Engine (options): %s (%s)\n",
 503         $t, $l, $e, $E, $s, $S);
 504     generateclips($l, $t, $e, $E, $s, $S);
 505     createvoice($l, $i);
 506     deletemp3s();
 507 }
 508 elsif ($C) {
 509     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);
 510     my $tts_object = init_tts($s, $S, $l);
 511     gentalkclips($ARGV[0], $tts_object, $e, $E, $S, 0);
 512     shutdown_tts($tts_object);
 513 }
 514 else {
 515     printusage();
 516     exit 1;
 517 }