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
  29 sub printusage {
  30     print <<USAGE
  31
  32 Usage: voice.pl [options] [path to dir]
  33  -V
  34     Create voice file. You must also specify -t and -l.
  35
  36  -C
  37     Create .talk clips.
  38
  39  -t=<target>
  40     Specify which target you want to build voicefile for. Must include
  41     any features that target supports.
  42
  43  -i=<target_id>
  44     Numeric target id. Needed for voice building.
  45
  46  -l=<language>
  47     Specify which language you want to build. Without .lang extension.
  48
  49  -e=<encoder>
  50     Which encoder to use for voice strings
  51
  52  -E=<encoder options>
  53     Which encoder options to use when compressing voice strings. Enclose
  54     in double quotes if the options include spaces.
  55
  56  -s=<TTS engine>
  57     Which TTS engine to use.
  58
  59  -S=<TTS engine options>
  60     Options to pass to the TTS engine. Enclose in double quotes if the
  61     options include spaces.
  62
  63  -v
  64     Be verbose
  65 USAGE
  66 ;
  67 }
  68
  69 # Initialize TTS engine. May return an object or value which will be passed
  70 # to voicestring and shutdown_tts
  71 sub init_tts {
  72     our $verbose;
  73     my ($tts_engine, $tts_engine_opts, $language) = @_;
  74     my %ret = ("name" => $tts_engine);
  75     switch($tts_engine) {
  76         case "festival" {
  77             print("> festival $tts_engine_opts --server\n") if $verbose;
  78             my $pid = open(FESTIVAL_SERVER, "| festival $tts_engine_opts --server > /dev/null 2>&1");
  79             my $dummy = *FESTIVAL_SERVER; #suppress warning
  80             $SIG{INT} = sub { kill TERM => $pid; print("foo"); panic_cleanup(); };
  81             $SIG{KILL} = sub { kill TERM => $pid; print("boo"); panic_cleanup(); };
  82             $ret{"pid"} = $pid;
  83         }
  84         case "sapi" {
  85             my $toolsdir = dirname($0);
  86             my $path = `cygpath $toolsdir -a -w`;
  87             chomp($path);
  88             $path = $path . '\\';
  89             my $cmd = $path . "sapi_voice.vbs /language:$language $tts_engine_opts";
  90             $cmd =~ s/\\/\\\\/g;
  91             print("> cscript //nologo $cmd\n") if $verbose;
  92             my $pid = open2(*CMD_OUT, *CMD_IN, "cscript //nologo $cmd");
  93             $SIG{INT} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
  94             $SIG{KILL} = sub { print(CMD_IN "QUIT\r\n"); panic_cleanup(); };
  95             print(CMD_IN "QUERY\tVENDOR\r\n");
  96             my $vendor = readline(CMD_OUT);
  97             $vendor =~ s/\r\n//;
  98             %ret = (%ret,
  99                     "stdin" => *CMD_IN,
 100                     "stdout" => *CMD_OUT,
 101                     "vendor" => $vendor);
 102         }
 103     }
 104     return \%ret;
 105 }
 106
 107 # Shutdown TTS engine if necessary.
 108 sub shutdown_tts {
 109     my ($tts_object) = @_;
 110     switch($$tts_object{"name"}) {
 111         case "festival" {
 112             # Send SIGTERM to festival server
 113             kill TERM => $$tts_object{"pid"};
 114         }
 115         case "sapi" {
 116             print({$$tts_object{"stdin"}} "QUIT\r\n");
 117             close($$tts_object{"stdin"});
 118         }
 119     }
 120 }
 121
 122 # Apply corrections to a voice-string to make it sound better
 123 sub correct_string {
 124     our $verbose;
 125     my ($string, $language, $tts_object) = @_;
 126     my $orig = $string;
 127     switch($language) {
 128         # General for all engines and languages
 129         $string =~ s/USB/U S B/g;
 130         $string =~ s/ID3/I D 3/g;
 131
 132         case "english" {
 133             switch($$tts_object{"name"}) {
 134                 case ["sapi","festival"] {
 135                     $string =~ s/plugin(s?)/plug-in$1/ig;
 136                 }
 137             }
 138         }
 139         case "deutsch" {
 140             # for all german engines (e.g. for english words)
 141             $string =~ s/alkaline/alkalein/ig;
 142             $string =~ s/byte(s?)/beit$1/ig;
 143             $string =~ s/clip(s?)/klipp$1/ig;
 144             $string =~ s/cuesheet/kjuschiet/ig;
 145             $string =~ s/dither/didder/ig;
 146             $string =~ s/equalizer/iquileiser/ig;
 147             $string =~ s/\bflash\b/fläsh/ig;
 148             $string =~ s/\bfirmware(s?)\b/firmwer$1/ig;
 149             $string =~ s/\bI D 3 tag\b/I D 3 täg/ig; # can't just use "tag" here
 150             $string =~ s/\bloudness\b/laudness/ig;
 151             $string =~ s/\bunicode\b/unikod/ig;
 152             switch($$tts_object{"name"}) {
 153                  case "sapi" {   # just for SAPI
 154                     switch($$tts_object{"vendor"}) {
 155                         case "AT&T Labs" {
 156                             $string =~ s/alphabet/alfabet/ig;
 157                             $string =~ s/ampere/amper/ig;
 158                             $string =~ s/\bdezibel\b/de-zibell/ig;
 159                             $string =~ s/diddering/didde-ring/ig;
 160                             $string =~ s/energie\b/ener-gie/ig;
 161                             $string =~ s/\bnumerisch\b/numehrisch/ig;
 162                             $string =~ s/\brücklauf\b/rück-lauf/ig;
 163                             $string =~ s/\bsuchlauf\b/such-lauf/ig;
 164                         }
 165                     }
 166                 }
 167             }
 168         }
 169         case "svenska" {
 170             # for all swedish engines (e.g. for english words)
 171             $string =~ s/kilobyte/kilobajt/ig;
 172             $string =~ s/megabyte/megabajt/ig;
 173             $string =~ s/gigabyte/gigabajt/ig;
 174             $string =~ s/\bloudness\b/laudness/ig;
 175
 176             switch($$tts_object{"name"}) {
 177                  case "espeak" {   # just for eSpeak
 178                      $string =~ s/ampere/ampär/ig;
 179                      $string =~ s/bokmärken/bok-märken/ig;
 180                      $string =~ s/generella/schenerella/ig;
 181                      $string =~ s/dithering/diddering/ig;
 182                      $string =~ s/\bunicode\b/jynikod/ig;
 183                      $string =~ s/uttoning/utoning/ig;
 184                      $string =~ s/procent/pro-cent/ig;
 185                      $string =~ s/spellistor/spelistor/ig;
 186                      $string =~ s/cuesheet/qjyschiit/ig;
 187                  }
 188             }
 189         }
 190         case "italiano" {
 191             # for all italian engines (e.g. for english words)
 192             $string =~ s/Replaygain/Ripleyghein/ig;
 193             $string =~ s/Crossfade/Crossfeid/ig;
 194             $string =~ s/beep/Bip/ig;
 195             $string =~ s/cuesheet/chiushit/ig;
 196             $string =~ s/fade/feid/ig;
 197             $string =~ s/Crossfeed/crossfid/ig;
 198             $string =~ s/Cache/chash/ig;
 199             $string =~ s/\bfirmware(s?)\b/firmuer$1/ig;
 200             $string =~ s/\bFile(s?)\b/fail$1/ig;
 201             $string =~ s/\bloudness\b/laudness/ig;
 202             $string =~ s/\bunicode\b/unikod/ig;
 203             $string =~ s/Playlist/pleylist/ig;
 204             $string =~ s/WavPack/wave pak/ig;
 205             $string =~ s/BITRATE/bit reit/ig;
 206             $string =~ s/Codepage/cod page/ig;
 207             $string =~ s/PCM Wave/pcm ueiv/ig;
 208             $string =~ s/Ã¨/è/ig;
 209             $string =~ s/\b(s*)Ã¹\b/$1ù/ig;
 210             $string =~ s/\b(s*)Ã\b/$1à/ig;
 211             switch($$tts_object{"name"}) {
 212                  case "sapi" {   # just for SAPI
 213                     switch($$tts_object{"vendor"}) {
 214                         case "Loquendo" {
 215                             $string =~ s/Inizializza/inizializa/ig;
 216                         }
 217                         case "ScanSoft, Inc" {
 218                             $string =~ s/V/v/ig;
 219                             $string =~ s/X/x/ig;
 220                             $string =~ s/stop/stohp/ig;
 221                         }
 222                     }
 223                 }
 224             }
 225         }
 226     }
 227     if ($orig ne $string) {
 228         printf("%s -> %s\n", $orig, $string) if $verbose;
 229     }
 230     return $string;
 231 }
 232
 233 # Produce a wav file of the text given
 234 sub voicestring {
 235     our $verbose;
 236     my ($string, $output, $tts_engine_opts, $tts_object) = @_;
 237     my $cmd;
 238     printf("Generate \"%s\" with %s in file %s\n", $string, $$tts_object{"name"}, $output) if $verbose;
 239     switch($$tts_object{"name"}) {
 240         case "festival" {
 241             # festival_client lies to us, so we have to do awful soul-eating
 242             # work with IPC::open3()
 243             $cmd = "festival_client --server localhost --otype riff --ttw --output \"$output\"";
 244             print("> $cmd\n") if $verbose;
 245             # Open command, and filehandles for STDIN, STDOUT, STDERR
 246             my $pid = open3(*CMD_IN, *CMD_OUT, *CMD_ERR, $cmd);
 247             # Put the string to speak into STDIN and close it
 248             print(CMD_IN $string);
 249             close(CMD_IN);
 250             # Read all output from festival_client (because it LIES TO US)
 251             while (<CMD_ERR>) {
 252             }
 253             close(CMD_OUT);
 254             close(CMD_ERR);
 255         }
 256         case "flite" {
 257             $cmd = "flite $tts_engine_opts -t \"$string\" \"$output\"";
 258             print("> $cmd\n") if $verbose;
 259             `$cmd`;
 260         }
 261         case "espeak" {
 262             # xxx: $tts_engine_opts isn't used
 263             $cmd = "espeak $tts_engine_opts -w \"$output\"";
 264             print("> $cmd\n") if $verbose;
 265             open(ESPEAK, "| $cmd");
 266             print ESPEAK $string . "\n";
 267             close(ESPEAK);
 268         }
 269         case "sapi" {
 270             print({$$tts_object{"stdin"}} "SPEAK\t$output\t$string\r\n");
 271         }
 272         case "swift" {
 273             $cmd = "swift $tts_engine_opts -o \"$output\" \"$string\"";
 274             print("> $cmd\n") if $verbose;
 275             system($cmd);
 276         }
 277     }
 278 }
 279
 280 # trim leading / trailing silence from the clip
 281 sub wavtrim {
 282     our $verbose;
 283     my ($file, $threshold, $tts_object) = @_;
 284     printf("Trim \"%s\"\n", $file) if $verbose;
 285     my $cmd = "wavtrim \"$file\" $threshold";
 286     if ($$tts_object{"name"} eq "sapi") {
 287         print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
 288     }
 289     else {
 290         print("> $cmd\n") if $verbose;
 291         `$cmd`;
 292     }
 293 }
 294
 295 # Encode a wav file into the given destination file
 296 sub encodewav {
 297     our $verbose;
 298     my ($input, $output, $encoder, $encoder_opts, $tts_object) = @_;
 299     printf("Encode \"%s\" with %s in file %s\n", $input, $encoder, $output) if $verbose;
 300     my $cmd = "$encoder $encoder_opts \"$input\" \"$output\"";
 301     if ($$tts_object{"name"} eq "sapi") {
 302         print({$$tts_object{"stdin"}} "EXEC\t$cmd\r\n");
 303     }
 304     else {
 305         print("> $cmd\n") if $verbose;
 306         `$cmd`;
 307     }
 308 }
 309
 310 # synchronize the clip generation / processing if it's running in another process
 311 sub synchronize {
 312     my ($tts_object) = @_;
 313     if ($$tts_object{"name"} eq "sapi") {
 314         print({$$tts_object{"stdin"}} "SYNC\t42\r\n");
 315         my $wait = readline($$tts_object{"stdout"});
 316         #ignore what's actually returned
 317     }
 318 }
 319
 320 # Run genlang and create voice clips for each string
 321 sub generateclips {
 322     our $verbose;
 323     my ($language, $target, $encoder, $encoder_opts, $tts_engine, $tts_engine_opts) = @_;
 324     my $english = dirname($0) . '/../apps/lang/english.lang';
 325     my $langfile = dirname($0) . '/../apps/lang/' . $language . '.lang';
 326     my $id = '';
 327     my $voice = '';
 328     my $cmd = "genlang -o -t=$target -e=$english $langfile 2>/dev/null";
 329     my $pool_file;
 330     open(VOICEFONTIDS, "> voicefontids");
 331     my $i = 0;
 332     local $| = 1; # make progress indicator work reliably
 333
 334     my $tts_object = init_tts($tts_engine, $tts_engine_opts, $language);
 335     print("Generating voice clips");
 336     print("\n") if $verbose;
 337     for (`$cmd`) {
 338         my $line = $_;
 339         print(VOICEFONTIDS $line);
 340         if ($line =~ /^id: (.*)$/) {
 341             $id = $1;
 342         }
 343         elsif ($line =~ /^voice: "(.*)"$/) {
 344             $voice = $1;
 345             if ($id !~ /^NOT_USED_.*$/ && $voice ne "") {
 346                 my $wav = $id . '.wav';
 347                 my $mp3 = $id . '.mp3';
 348
 349                 # Print some progress information
 350                 if (++$i % 10 == 0 and !$verbose) {
 351                     print(".");
 352                 }
 353
 354                 # Apply corrections to the string
 355                 $voice = correct_string($voice, $language, $tts_object);
 356
 357                 # If we have a pool of snippets, see if the string exists there first
 358                 if (defined($ENV{'POOL'})) {
 359                     $pool_file = sprintf("%s/%s-%s.mp3", $ENV{'POOL'},
 360                                          md5_hex("$voice $tts_engine $tts_engine_opts $encoder_opts"),
 361                                          $language);
 362                     if (-f $pool_file) {
 363                         printf("Re-using %s (%s) from pool\n", $id, $voice) if $verbose;
 364                         copy($pool_file, $mp3);
 365                     }
 366                 }
 367
 368                 # Don't generate MP3 if it already exists (probably from the POOL)
 369                 if (! -f $mp3) {
 370                     if ($id eq "VOICE_PAUSE") {
 371                         print("Use distributed $wav\n") if $verbose;
 372                         copy(dirname($0)."/VOICE_PAUSE.wav", $wav);
 373                     }
 374                     else {
 375                         voicestring($voice, $wav, $tts_engine_opts, $tts_object);
 376                         wavtrim($wav, 500, $tts_object);
 377                         # 500 seems to be a reasonable default for now
 378                     }
 379
 380                     encodewav($wav, $mp3, $encoder, $encoder_opts, $tts_object);
 381                     synchronize($tts_object);
 382                     if (defined($ENV{'POOL'})) {
 383                         copy($mp3, $pool_file);
 384                     }
 385                     unlink($wav);
 386                 }
 387                 $voice = "";
 388                 $id = "";
 389             }
 390         }
 391     }
 392     print("\n");
 393     close(VOICEFONTIDS);
 394     shutdown_tts($tts_object);
 395 }
 396
 397 # Assemble the voicefile
 398 sub createvoice {
 399     our $verbose;
 400     my ($language, $target_id) = @_;
 401     my $outfile = "";
 402     my $i = 0;
 403     do {
 404         $outfile = sprintf("%s%s.voice", $language, ($i++ == 0 ? '' : '-'.$i));
 405     } while (-f $outfile);
 406     printf("Saving voice file to %s\n", $outfile) if $verbose;
 407     my $cmd = "voicefont 'voicefontids' $target_id ./ $outfile";
 408     print("> $cmd\n") if $verbose;
 409     my $output = `$cmd`;
 410     print($output) if $verbose;
 411 }
 412
 413 sub deletemp3s() {
 414     for (glob('*.mp3')) {
 415         unlink($_);
 416     }
 417     for (glob('*.wav')) {
 418         unlink($_);
 419     }
 420 }
 421
 422 sub panic_cleanup {
 423     deletemp3s();
 424     die "moo";
 425 }
 426
 427 # Generate .talk clips
 428 sub gentalkclips {
 429     our $verbose;
 430     my ($dir, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i) = @_;
 431     my $d = new DirHandle $dir;
 432     while (my $file = $d->read) {
 433         my ($voice, $wav, $mp3);
 434         # Print some progress information
 435         if (++$i % 10 == 0 and !$verbose) {
 436             print(".");
 437         }
 438
 439         # Convert to a complete path
 440         my $path = sprintf("%s/%s", $dir, $file);
 441
 442         $voice = $file;
 443         $wav = sprintf("%s.talk.wav", $path);
 444
 445         # Ignore dot-dirs and talk files
 446         if ($file eq '.' || $file eq '..' || $file =~ /\.talk$/) {
 447             next;
 448         }
 449         # Element is a dir
 450         if ( -d $path) {
 451             gentalkclips($path, $tts_object, $encoder, $encoder_opts, $tts_engine_opts, $i);
 452             $mp3 = sprintf("%s/_dirname.talk", $path);
 453         }
 454         # Element is a file
 455         else {
 456             $mp3 = sprintf("%s.talk", $path);
 457             $voice =~ s/\.[^\.]*$//; # Trim extension
 458         }
 459
 460         printf("Talkclip %s: %s", $mp3, $voice) if $verbose;
 461
 462         voicestring($voice, $wav, $tts_engine_opts, $tts_object);
 463         wavtrim($wav, 500, $tts_object);
 464         # 500 seems to be a reasonable default for now
 465         encodewav($wav, $mp3, $encoder, $encoder_opts, $tts_object);
 466         unlink($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 }