/native/external/espeak/src/voices.cpp
C++ | 1707 lines | 1297 code | 309 blank | 101 comment | 321 complexity | 8f206aaa6a79bccd37babfabb96657b6 MD5 | raw file
1/*************************************************************************** 2 * Copyright (C) 2005 to 2007 by Jonathan Duddington * 3 * email: jonsd@users.sourceforge.net * 4 * * 5 * This program is free software; you can redistribute it and/or modify * 6 * it under the terms of the GNU General Public License as published by * 7 * the Free Software Foundation; either version 3 of the License, or * 8 * (at your option) any later version. * 9 * * 10 * This program is distributed in the hope that it will be useful, * 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of * 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * 13 * GNU General Public License for more details. * 14 * * 15 * You should have received a copy of the GNU General Public License * 16 * along with this program; if not, write see: * 17 * <http://www.gnu.org/licenses/>. * 18 ***************************************************************************/ 19 20#include "StdAfx.h" 21 22#include "stdio.h" 23#include "ctype.h" 24#include "wctype.h" 25#include "string.h" 26#include "stdlib.h" 27#include "speech.h" 28 29#ifdef PLATFORM_WINDOWS 30#include "windows.h" 31#else 32#ifdef PLATFORM_RISCOS 33#include "kernel.h" 34#else 35#include "dirent.h" 36#endif 37#endif 38 39#include "speak_lib.h" 40#include "phoneme.h" 41#include "synthesize.h" 42#include "voice.h" 43#include "translate.h" 44 45 46MNEM_TAB genders [] = { 47 {"unknown", 0}, 48 {"male", 1}, 49 {"female", 2}, 50 {NULL, 0 }}; 51 52//int tone_points[10] = {250,140, 1200,110, -1,0, -1,0, -1,0}; 53int tone_points[10] = {600,170, 1200,135, 2000,110, 3000,110, -1,0}; 54 55// limit the rate of change for each formant number 56//static int formant_rate_22050[9] = {50, 104, 165, 230, 220, 220, 220, 220, 220}; // values for 22kHz sample rate 57//static int formant_rate_22050[9] = {240, 180, 180, 180, 180, 180, 180, 180, 180}; // values for 22kHz sample rate 58static int formant_rate_22050[9] = {240, 170, 170, 170, 170, 170, 170, 170, 170}; // values for 22kHz sample rate 59int formant_rate[9]; // values adjusted for actual sample rate 60 61 62 63#define DEFAULT_LANGUAGE_PRIORITY 5 64#define N_VOICES_LIST 150 65static int n_voices_list = 0; 66static espeak_VOICE *voices_list[N_VOICES_LIST]; 67static int len_path_voices; 68 69espeak_VOICE voice_selected; 70 71 72 73#define V_NAME 1 74#define V_LANGUAGE 2 75#define V_GENDER 3 76#define V_TRANSLATOR 4 77#define V_PHONEMES 5 78#define V_DICTIONARY 6 79 80// these affect voice quality, are independent of language 81#define V_FORMANT 7 82#define V_PITCH 8 83#define V_ECHO 9 84#define V_FLUTTER 10 85#define V_ROUGHNESS 11 86#define V_CLARITY 12 87#define V_TONE 13 88#define V_VOICING 14 89#define V_BREATH 15 90#define V_BREATHW 16 91 92// these override defaults set by the translator 93#define V_WORDGAP 17 94#define V_INTONATION 18 95#define V_STRESSLENGTH 19 96#define V_STRESSAMP 20 97#define V_STRESSADD 21 98#define V_DICTRULES 22 99#define V_STRESSRULE 23 100#define V_CHARSET 24 101#define V_NUMBERS 25 102#define V_OPTION 26 103 104#define V_MBROLA 27 105 106// these need a phoneme table to have been specified 107#define V_REPLACE 28 108#define V_CONSONANTS 29 109 110 111typedef struct { 112 const char *mnem; 113 int data; 114} keywtab_t; 115 116static keywtab_t keyword_tab[] = { 117 {"name", V_NAME}, 118 {"language", V_LANGUAGE}, 119 {"gender", V_GENDER}, 120 121 {"formant", V_FORMANT}, 122 {"pitch", V_PITCH}, 123 {"phonemes", V_PHONEMES}, 124 {"translator", V_TRANSLATOR}, 125 {"dictionary", V_DICTIONARY}, 126 {"stressLength", V_STRESSLENGTH}, 127 {"stressAmp", V_STRESSAMP}, 128 {"stressAdd", V_STRESSADD}, 129 {"intonation", V_INTONATION}, 130 {"dictrules", V_DICTRULES}, 131 {"stressrule", V_STRESSRULE}, 132 {"charset", V_CHARSET}, 133 {"replace", V_REPLACE}, 134 {"words", V_WORDGAP}, 135 {"echo", V_ECHO}, 136 {"flutter", V_FLUTTER}, 137 {"roughness", V_ROUGHNESS}, 138 {"clarity", V_CLARITY}, 139 {"tone", V_TONE}, 140 {"voicing", V_VOICING}, 141 {"breath", V_BREATH}, 142 {"breathw", V_BREATHW}, 143 {"numbers", V_NUMBERS}, 144 {"option", V_OPTION}, 145 {"mbrola", V_MBROLA}, 146 {"consonants", V_CONSONANTS}, 147 148 // these just set a value in langopts.param[] 149 {"l_dieresis", 0x100+LOPT_DIERESES}, 150// {"l_lengthen", 0x100+LOPT_IT_LENGTHEN}, 151 {"l_prefix", 0x100+LOPT_PREFIXES}, 152 {"l_regressive_v", 0x100+LOPT_REGRESSIVE_VOICING}, 153 {"l_unpronouncable", 0x100+LOPT_UNPRONOUNCABLE}, 154 {"l_sonorant_min", 0x100+LOPT_SONORANT_MIN}, 155 {"l_length_mods", 0x100+LOPT_LENGTH_MODS}, 156 {NULL, 0} }; 157 158#define N_VOICE_VARIANTS 12 159const char variants_either[N_VOICE_VARIANTS] = {1,2,12,3,13,4,14,5,11,0}; 160const char variants_male[N_VOICE_VARIANTS] = {1,2,3,4,5,0}; 161const char variants_female[N_VOICE_VARIANTS] = {11,12,13,14,0}; 162const char *variant_lists[3] = {variants_either, variants_male, variants_female}; 163 164voice_t voicedata; 165voice_t *voice = &voicedata; 166 167char *fgets_strip(char *buf, int size, FILE *f_in) 168{//=============================================== 169// strip trailing spaces, and truncate lines at // comment 170 int len; 171 char *p; 172 173 if(fgets(buf,size,f_in) == NULL) 174 return(NULL); 175 176 len = strlen(buf); 177 while((--len > 0) && isspace(buf[len])) 178 buf[len] = 0; 179 180 if((p = strstr(buf,"//")) != NULL) 181 *p = 0; 182 183 return(buf); 184} 185 186 187void SetToneAdjust(voice_t *voice, int *tone_pts) 188{//============================================== 189 int ix; 190 int pt; 191 int y; 192 int freq1=0; 193 int freq2; 194 int height1 = tone_pts[1]; 195 int height2; 196 double rate; 197 198 for(pt=0; pt<10; pt+=2) 199 { 200 if(tone_pts[pt] == -1) 201 { 202 tone_pts[pt] = N_TONE_ADJUST*8; 203 if(pt > 0) 204 tone_pts[pt+1] = tone_pts[pt-1]; 205 } 206 freq2 = tone_pts[pt] / 8; // 8Hz steps 207 height2 = tone_pts[pt+1]; 208 if((freq2 - freq1) > 0) 209 { 210 rate = double(height2-height1)/(freq2-freq1); 211 212 for(ix=freq1; ix<freq2; ix++) 213 { 214 y = height1 + int(rate * (ix-freq1)); 215 if(y > 255) 216 y = 255; 217 voice->tone_adjust[ix] = y; 218 } 219 } 220 freq1 = freq2; 221 height1 = height2; 222 } 223} 224 225 226void ReadTonePoints(char *string, int *tone_pts) 227{//============================================= 228// tone_pts[] is int[10] 229 int ix; 230 231 for(ix=0; ix<10; ix++) 232 tone_pts[ix] = -1; 233 234 sscanf(string,"%d %d %d %d %d %d %d %d", 235 &tone_pts[0],&tone_pts[1],&tone_pts[2],&tone_pts[3], 236 &tone_pts[4],&tone_pts[5],&tone_pts[6],&tone_pts[7]); 237} 238 239 240 241 242static espeak_VOICE *ReadVoiceFile(FILE *f_in, const char *fname, const char*leafname) 243{//=================================================================================== 244// Read a Voice file, allocate a VOICE_DATA and set data from the 245// file's language, gender, name lines 246 247 char linebuf[120]; 248 char vname[80]; 249 char vgender[80]; 250 char vlanguage[80]; 251 char languages[300]; // allow space for several alternate language names and priorities 252 253 254 unsigned int len; 255 int langix = 0; 256 int n_languages = 0; 257 char *p; 258 espeak_VOICE *voice_data; 259 int priority; 260 int age; 261 int n_variants = 3; // default, number of variants of this voice before using another voice 262 int gender; 263 264#ifdef PLATFORM_WINDOWS 265 char fname_buf[sizeof(path_home)+15]; 266 if(memcmp(leafname,"mb-",3) == 0) 267 { 268 // check whether the mbrola speech data is present for this voice 269 memcpy(vname,&leafname[3],3); 270 vname[3] = 0; 271 sprintf(fname_buf,"%s/mbrola/%s",path_home,vname); 272 273 if(GetFileLength(fname_buf) <= 0) 274 return(0); 275 } 276#endif 277 278 vname[0] = 0; 279 vgender[0] = 0; 280 age = 0; 281 282 while(fgets_strip(linebuf,sizeof(linebuf),f_in) != NULL) 283 { 284 if(memcmp(linebuf,"name",4)==0) 285 { 286 p = &linebuf[4]; 287 while(isspace(*p)) p++; 288 strncpy0(vname,p,sizeof(vname)); 289 } 290 else 291 if(memcmp(linebuf,"language",8)==0) 292 { 293 priority = DEFAULT_LANGUAGE_PRIORITY; 294 vlanguage[0] = 0; 295 296 sscanf(&linebuf[8],"%s %d",vlanguage,&priority); 297 len = strlen(vlanguage) + 2; 298 // check for space in languages[] 299 if(len < (sizeof(languages)-langix-1)) 300 { 301 languages[langix] = priority; 302 303 strcpy(&languages[langix+1],vlanguage); 304 langix += len; 305 n_languages++; 306 } 307 } 308 else 309 if(memcmp(linebuf,"gender",6)==0) 310 { 311 sscanf(&linebuf[6],"%s %d",vgender,&age); 312 } 313 else 314 if(memcmp(linebuf,"variants",8)==0) 315 { 316 sscanf(&linebuf[8],"%d",&n_variants); 317 } 318 } 319 languages[langix++] = 0; 320 321 gender = LookupMnem(genders,vgender); 322 323 if(n_languages == 0) 324 { 325 return(NULL); // no language lines in the voice file 326 } 327 328 p = (char *)calloc(sizeof(espeak_VOICE) + langix + strlen(fname) + strlen(vname) + 3, 1); 329 voice_data = (espeak_VOICE *)p; 330 p = &p[sizeof(espeak_VOICE)]; 331 332 memcpy(p,languages,langix); 333 voice_data->languages = p; 334 335 strcpy(&p[langix],fname); 336 voice_data->identifier = &p[langix]; 337 voice_data->name = &p[langix]; 338 339 if(vname[0] != 0) 340 { 341 langix += strlen(fname)+1; 342 strcpy(&p[langix],vname); 343 voice_data->name = &p[langix]; 344 } 345 346 voice_data->age = age; 347 voice_data->gender = gender; 348 voice_data->variant = 0; 349 voice_data->xx1 = n_variants; 350 return(voice_data); 351} // end of ReadVoiceFile 352 353 354 355 356void VoiceReset(int tone_only) 357{//=========================== 358// Set voice to the default values 359 360 int pk; 361 static int breath_widths[N_PEAKS] = {0,200,200,400,400,400,600,600,600}; 362 363 // default is: pitch 82,118 364 voice->pitch_base = 0x49000; // default, 73 << 12; 365 voice->pitch_range = 0x0f30; // default = 0x1000 366 voice->formant_factor = 256; 367 368 voice->echo_delay = 0; 369 voice->echo_amp = 0; 370 voice->flutter = 64; 371 voice->n_harmonic_peaks = 5; 372 voice->peak_shape = 1; 373 voice->voicing = 64; 374 voice->consonant_amp = 100; 375 voice->consonant_ampv = 100; 376 377#ifdef PLATFORM_RISCOS 378 voice->roughness = 1; 379#else 380 voice->roughness = 2; 381#endif 382 383 InitBreath(); 384 for(pk=0; pk<N_PEAKS; pk++) 385 { 386 voice->freq[pk] = 256; 387 voice->height[pk] = 256; 388 voice->width[pk] = 256; 389 voice->breath[pk] = 0; 390 voice->breathw[pk] = breath_widths[pk]; // default breath formant woidths 391 voice->freqadd[pk] = 0; 392 393 // adjust formant smoothing depending on sample rate 394 formant_rate[pk] = (formant_rate_22050[pk] * 22050)/samplerate; 395 } 396 voice->height[2] = 240; // reduce F2 slightly 397 398 // This table provides the opportunity for tone control. 399 // Adjustment of harmonic amplitudes, steps of 8Hz 400 // value of 128 means no change 401// memset(voice->tone_adjust,128,sizeof(voice->tone_adjust)); 402SetToneAdjust(voice,tone_points); 403 404 // default values of speed factors 405 voice->speedf1 = 256; 406 voice->speedf2 = 238; 407 voice->speedf3 = 232; 408 409 if(tone_only == 0) 410 { 411 n_replace_phonemes = 0; 412 option_tone1 = 0; 413 option_quiet = 0; 414 LoadMbrolaTable(NULL,NULL,0); 415 } 416} // end of VoiceReset 417 418 419static void VoiceFormant(char *p) 420{//============================== 421 // Set parameters for a formant 422 int ix; 423 int formant; 424 int freq = 100; 425 int height = 100; 426 int width = 100; 427 int freqadd = 0; 428 429 ix = sscanf(p,"%d %d %d %d %d",&formant,&freq,&height,&width,&freqadd); 430 if(ix < 2) 431 return; 432 433 if((formant < 0) || (formant > 8)) 434 return; 435 436 if(freq >= 0) 437 voice->freq[formant] = int(freq * 2.56001); 438 if(height >= 0) 439 voice->height[formant] = int(height * 2.56001); 440 if(width >= 0) 441 voice->width[formant] = int(width * 2.56001); 442 voice->freqadd[formant] = freqadd; 443} 444 445 446 447 448 449static void PhonemeReplacement(int type, char *p) 450{//============================================== 451 int n; 452 int phon; 453 int flags = 0; 454 char phon_string1[12]; 455 char phon_string2[12]; 456 457 strcpy(phon_string2,"NULL"); 458 n = sscanf(p,"%d %s %s",&flags,phon_string1,phon_string2); 459 if((n < 2) || (n_replace_phonemes >= N_REPLACE_PHONEMES)) 460 return; 461 462 if((phon = LookupPh(phon_string1)) == 0) 463 return; // not recognised 464 465 replace_phonemes[n_replace_phonemes].old_ph = phon; 466 replace_phonemes[n_replace_phonemes].new_ph = LookupPh(phon_string2); 467 replace_phonemes[n_replace_phonemes++].type = flags; 468} // end of PhonemeReplacement 469 470 471 472static int Read8Numbers(char *data_in,int *data) 473{//============================================= 474// Read 8 integer numbers 475 return(sscanf(data_in,"%d %d %d %d %d %d %d %d", 476 &data[0],&data[1],&data[2],&data[3],&data[4],&data[5],&data[6],&data[7])); 477} 478 479 480voice_t *LoadVoice(const char *vname, int control) 481{//=============================================== 482// control, bit 0 1= no_default 483// bit 1 1 = change tone only, not language 484// bit 2 1 = don't report error on LoadDictionary 485// bit 4 1 = vname = full path 486 487 FILE *f_voice = NULL; 488 keywtab_t *k; 489 char *p; 490 int key; 491 int ix; 492 int n; 493 int value; 494 int error = 0; 495 int langix = 0; 496 int tone_only = control & 2; 497 int language_set = 0; 498 int phonemes_set = 0; 499 int stress_amps_set = 0; 500 int stress_lengths_set = 0; 501 int stress_add_set = 0; 502 int conditional_rules = 0; 503 LANGUAGE_OPTIONS *langopts = NULL; 504 505 Translator *new_translator = NULL; 506 507 char voicename[40]; 508 char language_name[40]; 509 char translator_name[40]; 510 char new_dictionary[40]; 511 char phonemes_name[40]; 512 const char *language_type; 513 char buf[200]; 514 char path_voices[sizeof(path_home)+12]; 515 char langname[4]; 516 517 int stress_amps[8]; 518 int stress_lengths[8]; 519 int stress_add[8]; 520 521 int pitch1; 522 int pitch2; 523 524 static char voice_identifier[40]; // file name for voice_selected 525 static char voice_name[40]; // voice name for voice_selected 526 static char voice_languages[100]; // list of languages and priorities for voice_selected 527 528 strcpy(voicename,vname); 529 if(voicename[0]==0) 530 strcpy(voicename,"default"); 531 532 if(control & 0x10) 533 { 534 strcpy(buf,vname); 535 if(GetFileLength(buf) <= 0) 536 return(NULL); 537 } 538 else 539 { 540 sprintf(path_voices,"%s%cvoices%c",path_home,PATHSEP,PATHSEP); 541 sprintf(buf,"%s%s",path_voices,voicename); 542 543 if(GetFileLength(buf) <= 0) 544 { 545 // look for the voice in a sub-directory of the language name 546 langname[0] = voicename[0]; 547 langname[1] = voicename[1]; 548 langname[2] = 0; 549 sprintf(buf,"%s%s%c%s",path_voices,langname,PATHSEP,voicename); 550 } 551 } 552 553 f_voice = fopen(buf,"r"); 554 555 language_type = "en"; // default 556 if(f_voice == NULL) 557 { 558 if(control & 3) 559 return(NULL); // can't open file 560 561 if(SelectPhonemeTableName(voicename) >= 0) 562 language_type = voicename; 563 } 564 565 if(!tone_only && (translator != NULL)) 566 { 567 delete translator; 568 translator = NULL; 569 } 570 571 strcpy(translator_name,language_type); 572 strcpy(new_dictionary,language_type); 573 strcpy(phonemes_name,language_type); 574 575 576 if(!tone_only) 577 { 578 voice = &voicedata; 579 strncpy0(voice_identifier,vname,sizeof(voice_identifier)); 580 voice_name[0] = 0; 581 voice_languages[0] = 0; 582 583 voice_selected.identifier = voice_identifier; 584 voice_selected.name = voice_name; 585 voice_selected.languages = voice_languages; 586 } 587 else 588 { 589 // append the variant file name to the voice identifier 590 if((p = strchr(voice_identifier,'+')) != NULL) 591 *p = 0; // remove previous variant name 592 sprintf(buf,"+%s",&vname[3]); // omit !v/ from the variant filename 593 strcat(voice_identifier,buf); 594 langopts = &translator->langopts; 595 } 596 VoiceReset(tone_only); 597 598 if(!tone_only) 599 SelectPhonemeTableName(phonemes_name); // set up phoneme_tab 600 601 602 while((f_voice != NULL) && (fgets_strip(buf,sizeof(buf),f_voice) != NULL)) 603 { 604 // isolate the attribute name 605 for(p=buf; (*p != 0) && !isspace(*p); p++); 606 *p++ = 0; 607 608 if(buf[0] == 0) continue; 609 610 key = 0; 611 for(k=keyword_tab; k->mnem != NULL; k++) 612 { 613 if(strcmp(buf,k->mnem)==0) 614 { 615 key = k->data; 616 break; 617 } 618 } 619 620 switch(key) 621 { 622 case V_LANGUAGE: 623 { 624 unsigned int len; 625 int priority; 626 627 if(tone_only) 628 break; 629 630 priority = DEFAULT_LANGUAGE_PRIORITY; 631 language_name[0] = 0; 632 633 sscanf(p,"%s %d",language_name,&priority); 634 if(strcmp(language_name,"variant") == 0) 635 break; 636 637 len = strlen(language_name) + 2; 638 // check for space in languages[] 639 if(len < (sizeof(voice_languages)-langix-1)) 640 { 641 voice_languages[langix] = priority; 642 643 strcpy(&voice_languages[langix+1],language_name); 644 langix += len; 645 } 646 647 // only act on the first language line 648 if(language_set == 0) 649 { 650 language_type = strtok(language_name,"-"); 651 language_set = 1; 652 strcpy(translator_name,language_type); 653 strcpy(new_dictionary,language_type); 654 strcpy(phonemes_name,language_type); 655 SelectPhonemeTableName(phonemes_name); 656 657 if(new_translator != NULL) 658 delete new_translator; 659 660 new_translator = SelectTranslator(translator_name); 661 langopts = &new_translator->langopts; 662 } 663 } 664 break; 665 666 case V_NAME: 667 if(tone_only == 0) 668 { 669 while(isspace(*p)) p++; 670 strncpy0(voice_name,p,sizeof(voice_name)); 671 } 672 break; 673 674 case V_GENDER: 675 { 676 int age; 677 char vgender[80]; 678 sscanf(p,"%s %d",vgender,&age); 679 voice_selected.gender = LookupMnem(genders,vgender); 680 voice_selected.age = age; 681 } 682 break; 683 684 case V_TRANSLATOR: 685 if(tone_only) break; 686 687 sscanf(p,"%s",translator_name); 688 689 if(new_translator != NULL) 690 delete new_translator; 691 692 new_translator = SelectTranslator(translator_name); 693 langopts = &new_translator->langopts; 694 break; 695 696 case V_DICTIONARY: // dictionary 697 sscanf(p,"%s",new_dictionary); 698 break; 699 700 case V_PHONEMES: // phoneme table 701 sscanf(p,"%s",phonemes_name); 702 break; 703 704 case V_FORMANT: 705 VoiceFormant(p); 706 break; 707 708 case V_PITCH: 709 { 710 double factor; 711 // default is pitch 82 118 712 n = sscanf(p,"%d %d",&pitch1,&pitch2); 713 voice->pitch_base = (pitch1 - 9) << 12; 714 voice->pitch_range = (pitch2 - pitch1) * 108; 715 factor = double(pitch1 - 82)/82; 716 voice->formant_factor = (int)((1+factor/4) * 256); // nominal formant shift for a different voice pitch 717 } 718 break; 719 720 case V_STRESSLENGTH: // stressLength 721 stress_lengths_set = Read8Numbers(p,stress_lengths); 722 break; 723 724 case V_STRESSAMP: // stressAmp 725 stress_amps_set = Read8Numbers(p,stress_amps); 726 break; 727 728 case V_STRESSADD: // stressAdd 729 stress_add_set = Read8Numbers(p,stress_add); 730 break; 731 732 case V_INTONATION: // intonation 733 sscanf(p,"%d %d",&option_tone_flags,&option_tone2); 734 if((option_tone_flags & 0xff) != 0) 735 langopts->intonation_group = option_tone_flags & 0xff; 736 break; 737 738 case V_DICTRULES: // conditional dictionary rules and list entries 739 while(*p != 0) 740 { 741 while(isspace(*p)) p++; 742 n = -1; 743 if(((n = atoi(p)) > 0) && (n < 32)) 744 { 745 p++; 746 conditional_rules |= (1 << n); 747 } 748 while(isalnum(*p)) p++; 749 } 750 break; 751 752 case V_REPLACE: 753 if(phonemes_set == 0) 754 { 755 // must set up a phoneme table before we can lookup phoneme mnemonics 756 SelectPhonemeTableName(phonemes_name); 757 phonemes_set = 1; 758 } 759 PhonemeReplacement(key,p); 760 break; 761 762 case V_WORDGAP: // words 763 sscanf(p,"%d %d",&langopts->word_gap, &langopts->vowel_pause); 764 break; 765 766 case V_STRESSRULE: 767 sscanf(p,"%d %d %d %d",&langopts->stress_rule, 768 &langopts->stress_flags, 769 &langopts->unstressed_wd1, 770 &langopts->unstressed_wd2); 771 break; 772 773 case V_CHARSET: 774 if((sscanf(p,"%d",&value)==1) && (value < N_CHARSETS)) 775 new_translator->charset_a0 = charsets[value]; 776 break; 777 778 case V_NUMBERS: 779 sscanf(p,"%d",&langopts->numbers); 780 break; 781 782 case V_OPTION: 783 if(sscanf(p,"%d %d",&ix,&value) == 2) 784 { 785 if((ix >= 0) && (ix < N_LOPTS)) 786 langopts->param[ix] = value; 787 } 788 break; 789 790 case V_ECHO: 791 // echo. suggest: 135mS 11% 792 value = 0; 793 voice->echo_amp = 0; 794 sscanf(p,"%d %d",&voice->echo_delay,&voice->echo_amp); 795 break; 796 797 case V_FLUTTER: // flutter 798 if(sscanf(p,"%d",&value)==1) 799 voice->flutter = value * 32; 800 break; 801 802 case V_ROUGHNESS: // roughness 803 if(sscanf(p,"%d",&value)==1) 804 voice->roughness = value; 805 break; 806 807 case V_CLARITY: // formantshape 808 if(sscanf(p,"%d",&value)==1) 809 { 810 if(value > 4) 811 { 812 voice->peak_shape = 1; // squarer formant peaks 813 value = 4; 814 } 815 voice->n_harmonic_peaks = 1+value; 816 } 817 break; 818 819 case V_TONE: 820 { 821 int tone_data[10]; 822 ReadTonePoints(p,tone_data); 823 SetToneAdjust(voice,tone_data); 824 } 825 break; 826 827 case V_VOICING: 828 if(sscanf(p,"%d",&value)==1) 829 voice->voicing = (value * 64)/100; 830 break; 831 832 case V_BREATH: 833 voice->breath[0] = Read8Numbers(p,&voice->breath[1]); 834 for(ix=1; ix<8; ix++) 835 { 836 if(ix % 2) 837 voice->breath[ix] = -voice->breath[ix]; 838 } 839 break; 840 841 case V_BREATHW: 842 voice->breathw[0] = Read8Numbers(p,&voice->breathw[1]); 843 break; 844 845 case V_CONSONANTS: 846 value = sscanf(p,"%d %d",&voice->consonant_amp, &voice->consonant_ampv); 847 break; 848 849 case V_MBROLA: 850 { 851 int srate = 16000; 852 char name[40]; 853 char phtrans[40]; 854 855 phtrans[0] = 0; 856 sscanf(p,"%s %s %d",name,phtrans,&srate); 857 LoadMbrolaTable(name,phtrans,srate); 858 } 859 break; 860 861 default: 862 if((key & 0xff00) == 0x100) 863 { 864 sscanf(p,"%d",&langopts->param[key &0xff]); 865 } 866 else 867 { 868 fprintf(stderr,"Bad voice attribute: %s\n",buf); 869 } 870 break; 871 } 872 } 873 if(f_voice != NULL) 874 fclose(f_voice); 875 876 if((new_translator == NULL) && (!tone_only)) 877 { 878 // not set by language attribute 879 new_translator = SelectTranslator(translator_name); 880 } 881 882 for(ix=0; ix<N_PEAKS; ix++) 883 { 884 voice->freq2[ix] = voice->freq[ix]; 885 voice->height2[ix] = voice->height[ix]; 886 voice->width2[ix] = voice->width[ix]; 887 } 888 889 if(tone_only) 890 { 891 new_translator = translator; 892 } 893 else 894 { 895 if((ix = SelectPhonemeTableName(phonemes_name)) < 0) 896 { 897 fprintf(stderr,"Unknown phoneme table: '%s'\n",phonemes_name); 898 } 899 voice->phoneme_tab_ix = ix; 900 error = new_translator->LoadDictionary(new_dictionary, control & 4); 901 if(dictionary_name[0]==0) 902 return(NULL); // no dictionary loaded 903 904 new_translator->dict_condition = conditional_rules; 905 906 voice_languages[langix] = 0; 907 } 908 909 langopts = &new_translator->langopts; 910 911 912 if((value = langopts->param[LOPT_LENGTH_MODS]) != 0) 913 { 914 SetLengthMods(new_translator,value); 915 } 916 917 voice->width[0] = (voice->width[0] * 105)/100; 918 919 if(!tone_only) 920 { 921 translator = new_translator; 922 } 923 924 // relative lengths of different stress syllables 925 for(ix=0; ix<stress_lengths_set; ix++) 926 { 927 translator->stress_lengths[ix] = stress_lengths[ix]; 928 } 929 for(ix=0; ix<stress_add_set; ix++) 930 { 931 translator->stress_lengths[ix] += stress_add[ix]; 932 } 933 for(ix=0; ix<stress_amps_set; ix++) 934 { 935 translator->stress_amps[ix] = stress_amps[ix]; 936 translator->stress_amps_r[ix] = stress_amps[ix] -1; 937 } 938 939 return(voice); 940} // end of LoadVoice 941 942 943char *ExtractVoiceVariantName(char *vname, int variant_num) 944{//======================================================== 945// Remove any voice variant suffix (name or number) from a voice name 946// Returns the voice variant name 947 948 char *p; 949 static char variant_name[20]; 950 char variant_prefix[5]; 951 952 variant_name[0] = 0; 953 sprintf(variant_prefix,"!v%c",PATHSEP); 954 955 if(vname != NULL) 956 { 957 if((p = strchr(vname,'+')) != NULL) 958 { 959 // The voice name has a +variant suffix 960 *p++ = 0; // delete the suffix from the voice name 961 if(isdigit(*p)) 962 { 963 variant_num = atoi(p); // variant number 964 } 965 else 966 { 967 // voice variant name, not number 968 strcpy(variant_name,variant_prefix); 969 strncpy0(&variant_name[3],p,sizeof(variant_name)-3); 970 } 971 } 972 } 973 974 if(variant_num > 0) 975 { 976 if(variant_num < 10) 977 sprintf(variant_name,"%sm%d",variant_prefix, variant_num); // male 978 else 979 sprintf(variant_name,"%sf%d",variant_prefix, variant_num-10); // female 980 } 981 982 return(variant_name); 983} // end of ExtractVoiceVariantName 984 985 986 987voice_t *LoadVoiceVariant(const char *vname, int variant_num) 988{//========================================================== 989// Load a voice file. 990// Also apply a voice variant if specified by "variant", or by "+number" or "+name" in the "vname" 991 992 voice_t *v; 993 char *variant_name; 994 char buf[60]; 995 996 strncpy0(buf,vname,sizeof(buf)); 997 variant_name = ExtractVoiceVariantName(buf,variant_num); 998 999 if((v = LoadVoice(buf,0)) == NULL) 1000 return(NULL); 1001 1002 if(variant_name[0] != 0) 1003 { 1004 v = LoadVoice(variant_name,2); 1005 } 1006 return(v); 1007} 1008 1009 1010 1011static int __cdecl VoiceNameSorter(const void *p1, const void *p2) 1012{//======================================================= 1013 int ix; 1014 espeak_VOICE *v1 = *(espeak_VOICE **)p1; 1015 espeak_VOICE *v2 = *(espeak_VOICE **)p2; 1016 1017 1018 if((ix = strcmp(&v1->languages[1],&v2->languages[1])) != 0) // primary language name 1019 return(ix); 1020 if((ix = v1->languages[0] - v2->languages[0]) != 0) // priority number 1021 return(ix); 1022 return(strcmp(v1->name,v2->name)); 1023} 1024 1025 1026static int __cdecl VoiceScoreSorter(const void *p1, const void *p2) 1027{//======================================================== 1028 int ix; 1029 espeak_VOICE *v1 = *(espeak_VOICE **)p1; 1030 espeak_VOICE *v2 = *(espeak_VOICE **)p2; 1031 1032 if((ix = v2->score - v1->score) != 0) 1033 return(ix); 1034 return(strcmp(v1->name,v2->name)); 1035} 1036 1037 1038static int ScoreVoice(espeak_VOICE *voice_spec, const char *spec_language, int spec_n_parts, int spec_lang_len, espeak_VOICE *voice) 1039{//========================================================================================================================= 1040 int ix; 1041 const char *p; 1042 int c1, c2; 1043 int language_priority; 1044 int n_parts; 1045 int matching; 1046 int matching_parts; 1047 int score = 0; 1048 int x; 1049 int ratio; 1050 int required_age; 1051 int diff; 1052 1053 p = voice->languages; // list of languages+dialects for which this voice is suitable 1054 1055 if(strcmp(spec_language,"mbrola")==0) 1056 { 1057 // only list mbrola voices 1058 if(memcmp(voice->identifier,"mb/",3) == 0) 1059 return(100); 1060 return(0); 1061 } 1062 1063 if(spec_n_parts == 0) 1064 { 1065 score = 100; 1066 } 1067 else 1068 { 1069 if((*p == 0) && (strcmp(spec_language,"variants")==0)) 1070 { 1071 // match on a voice with no languages if the required language is "variants" 1072 score = 100; 1073 } 1074 1075 // compare the required language with each of the languages of this voice 1076 while(*p != 0) 1077 { 1078 language_priority = *p++; 1079 1080 matching = 1; 1081 matching_parts = 0; 1082 n_parts = 1; 1083 1084 for(ix=0; ; ix++) 1085 { 1086 if((ix >= spec_lang_len) || ((c1 = spec_language[ix]) == '-')) 1087 c1 = 0; 1088 if((c2 = p[ix]) == '-') 1089 c2 = 0; 1090 1091 if(c1 != c2) 1092 { 1093 matching = 0; 1094 } 1095 1096 if(p[ix] == '-') 1097 { 1098 n_parts++; 1099 if(matching) 1100 matching_parts++; 1101 } 1102 if(p[ix] == 0) 1103 break; 1104 } 1105 p += (ix+1); 1106 matching_parts += matching; // number of parts which match 1107 1108 if(matching_parts == 0) 1109 break; // no matching parts for this language 1110 1111 x = 5; 1112 // reduce the score if not all parts of the required language match 1113 if((diff = (spec_n_parts - matching_parts)) > 0) 1114 x -= diff; 1115 1116 // reduce score if the language is more specific than required 1117 if((diff = (n_parts - matching_parts)) > 0) 1118 x -= diff; 1119 1120 x = x*100 - (language_priority * 2); 1121 1122 if(x > score) 1123 score = x; 1124 } 1125 } 1126 if(score == 0) 1127 return(0); 1128 1129 if(voice_spec->name != NULL) 1130 { 1131 if(strcmp(voice_spec->name,voice->name)==0) 1132 { 1133 // match on voice name 1134 score += 500; 1135 } 1136 else 1137 if(strcmp(voice_spec->name,voice->identifier)==0) 1138 { 1139 score += 400; 1140 } 1141 } 1142 1143 if(((voice_spec->gender == 1) || (voice_spec->gender == 2)) && 1144 ((voice->gender == 1) || (voice->gender == 2))) 1145 { 1146 if(voice_spec->gender == voice->gender) 1147 score += 50; 1148 else 1149 score -= 50; 1150 } 1151 1152 if((voice_spec->age <= 12) && (voice->gender == 2) && (voice->age > 12)) 1153 { 1154 score += 5; // give some preference for non-child female voice if a child is requested 1155 } 1156 1157 if(voice->age != 0) 1158 { 1159 if(voice_spec->age == 0) 1160 required_age = 30; 1161 else 1162 required_age = voice_spec->age; 1163 1164 ratio = (required_age*100)/voice->age; 1165 if(ratio < 100) 1166 ratio = 10000/ratio; 1167 ratio = (ratio - 100)/10; // 0=exact match, 10=out by factor of 2 1168 x = 5 - ratio; 1169 if(x > 0) x = 0; 1170 1171 score = score + x; 1172 1173 if(voice_spec->age > 0) 1174 score += 10; // required age specified, favour voices with a specified age (near it) 1175 } 1176 if(score < 1) 1177 score = 1; 1178 return(score); 1179} // end of ScoreVoice 1180 1181 1182static int SetVoiceScores(espeak_VOICE *voice_select, espeak_VOICE **voices, int control) 1183{//====================================================================================== 1184// control: bit0=1 include mbrola voices 1185 int ix; 1186 int score; 1187 int nv; // number of candidates 1188 int n_parts=0; 1189 int lang_len=0; 1190 espeak_VOICE *vp; 1191 char language[80]; 1192 1193 // count number of parts in the specified language 1194 if((voice_select->languages != NULL) && (voice_select->languages[0] != 0)) 1195 { 1196 n_parts = 1; 1197 lang_len = strlen(voice_select->languages); 1198 for(ix=0; (ix<=lang_len) && ((unsigned)ix < sizeof(language)); ix++) 1199 { 1200 if((language[ix] = tolower(voice_select->languages[ix])) == '-') 1201 n_parts++; 1202 } 1203 } 1204 // select those voices which match the specified language 1205 nv = 0; 1206 for(ix=0; ix<n_voices_list; ix++) 1207 { 1208 vp = voices_list[ix]; 1209 1210 if(((control & 1) == 0) && (memcmp(vp->identifier,"mb/",3) == 0)) 1211 continue; 1212 1213 if((score = ScoreVoice(voice_select, language, n_parts, lang_len, voices_list[ix])) > 0) 1214 { 1215 voices[nv++] = vp; 1216 vp->score = score; 1217 } 1218 } 1219 voices[nv] = NULL; // list terminator 1220 1221 if(nv==0) 1222 return(0); 1223 1224 // sort the selected voices by their score 1225 qsort(voices,nv,sizeof(espeak_VOICE *),(int (__cdecl *)(const void *,const void *))VoiceScoreSorter); 1226 1227 return(nv); 1228} // end of SetVoiceScores 1229 1230 1231 1232 1233espeak_VOICE *SelectVoiceByName(espeak_VOICE **voices, const char *name) 1234{//===================================================================== 1235 int ix; 1236 int match_fname = -1; 1237 int match_fname2 = -1; 1238 int match_name = -1; 1239 const char *id; 1240 int last_part_len; 1241 char last_part[41]; 1242 1243 if(voices == NULL) 1244 { 1245 if(n_voices_list == 0) 1246 espeak_ListVoices(NULL); // create the voices list 1247 voices = voices_list; 1248 } 1249 1250 sprintf(last_part,"%c%s",PATHSEP,name); 1251 last_part_len = strlen(last_part); 1252 1253 for(ix=0; voices[ix] != NULL; ix++) 1254 { 1255 if(strcmp(name,voices[ix]->name)==0) 1256 { 1257 match_name = ix; // found matching voice name 1258 break; 1259 } 1260 else 1261 if(strcmp(name,id = voices[ix]->identifier)==0) 1262 { 1263 match_fname = ix; // matching identifier, use this if no matching name 1264 } 1265 else 1266 if(strcmp(last_part,&id[strlen(id)-last_part_len])==0) 1267 { 1268 match_fname2 = ix; 1269 } 1270 } 1271 1272 if(match_name < 0) 1273 { 1274 match_name = match_fname; // no matching name, try matching filename 1275 if(match_name < 0) 1276 match_name = match_fname2; // try matching just the last part of the filename 1277 } 1278 1279 if(match_name < 0) 1280 return(NULL); 1281 1282 return(voices[match_name]); 1283} // end of SelectVoiceByName 1284 1285 1286 1287 1288char const *SelectVoice(espeak_VOICE *voice_select) 1289{//================================================ 1290// Returns a path within espeak-voices, with a possible +variant suffix 1291// variant is an output-only parameter 1292 int nv; // number of candidates 1293 int ix, ix2; 1294 int j; 1295 int n_variants; 1296 int variant_number; 1297 int gender; 1298 int skip; 1299 int aged=1; 1300 char *variant_name; 1301 const char *p, *p_start; 1302 espeak_VOICE *vp = NULL; 1303 espeak_VOICE *vp2; 1304 espeak_VOICE voice_select2; 1305 espeak_VOICE *voices[N_VOICES_LIST]; // list of candidates 1306 espeak_VOICE *voices2[N_VOICES_LIST+N_VOICE_VARIANTS]; 1307 static espeak_VOICE voice_variants[N_VOICE_VARIANTS]; 1308 static char voice_id[50]; 1309 1310 memcpy(&voice_select2,voice_select,sizeof(voice_select2)); 1311 1312 if(n_voices_list == 0) 1313 espeak_ListVoices(NULL); // create the voices list 1314 1315 if((voice_select2.languages == NULL) || (voice_select2.languages[0] == 0)) 1316 { 1317 // no language is specified. Get language from the named voice 1318 static char buf[60]; 1319 1320 if(voice_select2.name == NULL) 1321 { 1322 if((voice_select2.name = voice_select2.identifier) == NULL) 1323 voice_select2.name = "default"; 1324 } 1325 1326 strncpy0(buf,voice_select2.name,sizeof(buf)); 1327 variant_name = ExtractVoiceVariantName(buf,0); 1328 1329 vp = SelectVoiceByName(voices_list,buf); 1330 if(vp != NULL) 1331 { 1332 voice_select2.languages = &(vp->languages[1]); 1333 1334 if((voice_select2.gender==0) && (voice_select2.age==0) && (voice_select2.variant==0)) 1335 { 1336 if(variant_name[0] != 0) 1337 { 1338 sprintf(voice_id,"%s+%s",vp->identifier,&variant_name[3]); // omit the !v/ from variant_name 1339 return(voice_id); 1340 } 1341 1342 return(vp->identifier); 1343 } 1344 } 1345 } 1346 1347 // select and sort voices for the required language 1348 nv = SetVoiceScores(&voice_select2,voices,0); 1349 1350 if(nv == 0) 1351 { 1352 // no matching voice, choose the default 1353 if((voices[0] = SelectVoiceByName(voices_list,"default")) != NULL) 1354 nv = 1; 1355 } 1356 1357 gender = 0; 1358 if((voice_select2.gender == 2) || ((voice_select2.age > 0) && (voice_select2.age < 13))) 1359 gender = 2; 1360 else 1361 if(voice_select2.gender == 1) 1362 gender = 1; 1363 1364#define AGE_OLD 60 1365 if(voice_select2.age < AGE_OLD) 1366 aged = 0; 1367 1368 p = p_start = variant_lists[gender]; 1369 if(aged == 0) 1370 p++; // the first voice in the variants list is older 1371 1372 // add variants for the top voices 1373 n_variants = 0; 1374 for(ix=0, ix2=0; ix<nv; ix++) 1375 { 1376 vp = voices[ix]; 1377 // is the main voice the required gender? 1378 skip=0; 1379 if((gender != 0) && (vp->gender != gender)) 1380 { 1381 skip=1; 1382 } 1383 if((ix2==0) && aged && (vp->age < AGE_OLD)) 1384 { 1385 skip=1; 1386 } 1387 if(skip==0) 1388 { 1389 voices2[ix2++] = vp; 1390 } 1391 1392 for(j=0; (j < vp->xx1) && (n_variants < N_VOICE_VARIANTS);) 1393 { 1394 if((variant_number = *p) == 0) 1395 { 1396 p = p_start; 1397 continue; 1398 } 1399 1400 vp2 = &voice_variants[n_variants++]; // allocate space for voice variant 1401 memcpy(vp2,vp,sizeof(espeak_VOICE)); // copy from the original voice 1402 vp2->variant = variant_number; 1403 voices2[ix2++] = vp2; 1404 p++; 1405 j++; 1406 } 1407 } 1408 // add any more variants to the end of the list 1409 while((vp != NULL) && ((variant_number = *p++) != 0) && (n_variants < N_VOICE_VARIANTS)) 1410 { 1411 vp2 = &voice_variants[n_variants++]; // allocate space for voice variant 1412 memcpy(vp2,vp,sizeof(espeak_VOICE)); // copy from the original voice 1413 vp2->variant = variant_number; 1414 voices2[ix2++] = vp2; 1415 } 1416 1417 // index the sorted list by the required variant number 1418 vp = voices2[voice_select2.variant % ix2]; 1419 1420 if(vp->variant != 0) 1421 { 1422 variant_name = ExtractVoiceVariantName(NULL,vp->variant); 1423 sprintf(voice_id,"%s+%s",vp->identifier,&variant_name[3]); 1424 return(voice_id); 1425 } 1426 1427 return(vp->identifier); 1428} // end of SelectVoice 1429 1430 1431 1432void GetVoices(const char *path) 1433{//============================= 1434 FILE *f_voice; 1435 espeak_VOICE *voice_data; 1436 int ftype; 1437 char fname[sizeof(path_home)+100]; 1438 1439#ifdef PLATFORM_RISCOS 1440 int len; 1441 int *type; 1442 char *p; 1443 _kernel_swi_regs regs; 1444 _kernel_oserror *error; 1445 char buf[80]; 1446 char directory2[sizeof(path_home)+100]; 1447 1448 regs.r[0] = 10; 1449 regs.r[1] = (int)path; 1450 regs.r[2] = (int)buf; 1451 regs.r[3] = 1; 1452 regs.r[4] = 0; 1453 regs.r[5] = sizeof(buf); 1454 regs.r[6] = 0; 1455 1456 while(regs.r[3] > 0) 1457 { 1458 error = _kernel_swi(0x0c+0x20000,®s,®s); /* OS_GBPB 10, read directory entries */ 1459 if((error != NULL) || (regs.r[3] == 0)) 1460 { 1461 break; 1462 } 1463 type = (int *)(&buf[16]); 1464 len = strlen(&buf[20]); 1465 sprintf(fname,"%s.%s",path,&buf[20]); 1466 1467 if(*type == 2) 1468 { 1469 // a sub-directory 1470 GetVoices(fname); 1471 } 1472 else 1473 { 1474 // a regular line, add it to the voices list 1475 if((f_voice = fopen(fname,"r")) == NULL) 1476 continue; 1477 1478 // pass voice file name within the voices directory 1479 voice_data = ReadVoiceFile(f_voice, fname+len_path_voices, &buf[20]); 1480 fclose(f_voice); 1481 1482 if(voice_data != NULL) 1483 { 1484 voices_list[n_voices_list++] = voice_data; 1485 } 1486 } 1487 } 1488#else 1489#ifdef PLATFORM_WINDOWS 1490 WIN32_FIND_DATA FindFileData; 1491 HANDLE hFind = INVALID_HANDLE_VALUE; 1492 DWORD dwError; 1493 1494#undef UNICODE // we need FindFirstFileA() which takes an 8-bit c-string 1495 sprintf(fname,"%s\\*",path); 1496 hFind = FindFirstFile(fname, &FindFileData); 1497 if(hFind == INVALID_HANDLE_VALUE) 1498 return; 1499 1500 do { 1501 sprintf(fname,"%s%c%s",path,PATHSEP,FindFileData.cFileName); 1502 1503 ftype = GetFileLength(fname); 1504 1505 if((ftype == -2) && (FindFileData.cFileName[0] != '.')) 1506 { 1507 // a sub-sirectory 1508 GetVoices(fname); 1509 } 1510 else 1511 if(ftype > 0) 1512 { 1513 // a regular line, add it to the voices list 1514 if((f_voice = fopen(fname,"r")) == NULL) 1515 continue; 1516 1517 // pass voice file name within the voices directory 1518 voice_data = ReadVoiceFile(f_voice, fname+len_path_voices, FindFileData.cFileName); 1519 fclose(f_voice); 1520 1521 if(voice_data != NULL) 1522 { 1523 voices_list[n_voices_list++] = voice_data; 1524 } 1525 } 1526 } while(FindNextFile(hFind, &FindFileData) != 0); 1527 FindClose(hFind); 1528 1529#else 1530 DIR *dir; 1531 struct dirent *ent; 1532 1533 if((dir = opendir(path)) == NULL) 1534 return; 1535 1536 while((ent = readdir(dir)) != NULL) 1537 { 1538 if(n_voices_list >= (N_VOICES_LIST-2)) 1539 break; // voices list is full 1540 1541 sprintf(fname,"%s%c%s",path,PATHSEP,ent->d_name); 1542 1543 ftype = GetFileLength(fname); 1544 1545 if((ftype == -2) && (ent->d_name[0] != '.')) 1546 { 1547 // a sub-sirectory 1548 GetVoices(fname); 1549 } 1550 else 1551 if(ftype > 0) 1552 { 1553 // a regular line, add it to the voices list 1554 if((f_voice = fopen(fname,"r")) == NULL) 1555 continue; 1556 1557 // pass voice file name within the voices directory 1558 voice_data = ReadVoiceFile(f_voice, fname+len_path_voices, ent->d_name); 1559 fclose(f_voice); 1560 1561 if(voice_data != NULL) 1562 { 1563 voices_list[n_voices_list++] = voice_data; 1564 } 1565 } 1566 } 1567 closedir(dir); 1568#endif 1569#endif 1570} // end of GetVoices 1571 1572 1573 1574espeak_ERROR SetVoiceByName(const char *name) 1575{//========================================= 1576 espeak_VOICE *v; 1577 espeak_VOICE voice_selector; 1578 char *variant_name; 1579 static char buf[60]; 1580 1581 strncpy0(buf,name,sizeof(buf)); 1582 variant_name = ExtractVoiceVariantName(buf,0); 1583 1584 memset(&voice_selector,0,sizeof(voice_selector)); 1585// voice_selector.name = buf; 1586 voice_selector.name = (char *)name; // include variant name in voice stack ?? 1587 1588 // first check for a voice with this filename 1589 // This may avoid the need to call espeak_ListVoices(). 1590 1591 if(LoadVoice(buf,1) != NULL) 1592 { 1593 if(variant_name[0] != 0) 1594 { 1595 LoadVoice(variant_name,2); 1596 } 1597 1598 DoVoiceChange(voice); 1599 SetVoiceStack(&voice_selector); 1600 return(EE_OK); 1601 } 1602 1603 if(n_voices_list == 0) 1604 espeak_ListVoices(NULL); // create the voices list 1605 1606 if((v = SelectVoiceByName(voices_list,buf)) != NULL) 1607 { 1608 if(LoadVoice(v->identifier,0) != NULL) 1609 { 1610 if(variant_name[0] != 0) 1611 { 1612 LoadVoice(variant_name,2); 1613 } 1614 DoVoiceChange(voice); 1615 SetVoiceStack(&voice_selector); 1616 return(EE_OK); 1617 } 1618 } 1619 return(EE_INTERNAL_ERROR); // voice name not found 1620} // end of SetVoiceByName 1621 1622 1623 1624espeak_ERROR SetVoiceByProperties(espeak_VOICE *voice_selector) 1625{//============================================================ 1626 const char *voice_id; 1627 1628 voice_id = SelectVoice(voice_selector); 1629 1630 LoadVoiceVariant(voice_id,0); 1631 DoVoiceChange(voice); 1632 SetVoiceStack(voice_selector); 1633 1634 return(EE_OK); 1635} // end of SetVoiceByProperties 1636 1637 1638 1639 1640//======================================================================= 1641// Library Interface Functions 1642//======================================================================= 1643#pragma GCC visibility push(default) 1644 1645 1646ESPEAK_API const espeak_VOICE **espeak_ListVoices(espeak_VOICE *voice_spec) 1647{//======================================================================== 1648#ifndef PLATFORM_RISCOS 1649 int ix; 1650 int j; 1651 espeak_VOICE *v; 1652 static espeak_VOICE *voices[N_VOICES_LIST]; 1653 char path_voices[sizeof(path_home)+12]; 1654 1655 // free previous voice list data 1656 1657 for(ix=0; ix<n_voices_list; ix++) 1658 { 1659 if(voices_list[ix] != NULL) 1660 free(voices_list[ix]); 1661 } 1662 n_voices_list = 0; 1663 1664 sprintf(path_voices,"%s%cvoices",path_home,PATHSEP); 1665 len_path_voices = strlen(path_voices)+1; 1666 1667 GetVoices(path_voices); 1668 voices_list[n_voices_list] = NULL; // voices list terminator 1669 1670 // sort the voices list 1671 qsort(voices_list,n_voices_list,sizeof(espeak_VOICE *), 1672 (int (__cdecl *)(const void *,const void *))VoiceNameSorter); 1673 1674 1675 if(voice_spec) 1676 { 1677 // select the voices which match the voice_spec, and sort them by preference 1678 SetVoiceScores(voice_spec,voices,1); 1679 } 1680 else 1681 { 1682 // list all: omit variant voices and mbrola voices 1683 j = 0; 1684 for(ix=0; (v = voices_list[ix]) != NULL; ix++) 1685 { 1686 if((v->languages[0] != 0) && (strcmp(&v->languages[1],"variant") != 0) && (memcmp(v->identifier,"mb/",3) != 0)) 1687 { 1688 voices[j++] = v; 1689 } 1690 } 1691 voices[j] = NULL; 1692 } 1693 return((const espeak_VOICE **)voices); 1694#endif 1695 return((const espeak_VOICE **)voices_list); 1696} // end of espeak_ListVoices 1697 1698 1699 1700ESPEAK_API espeak_VOICE *espeak_GetCurrentVoice(void) 1701{//================================================== 1702 return(&voice_selected); 1703} 1704 1705#pragma GCC visibility pop 1706 1707