/native/external/espeak/src/synthesize.cpp
C++ | 1628 lines | 1247 code | 287 blank | 94 comment | 292 complexity | 9f3a10693542d29d1e0f1ec7b3f9fc32 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, 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 <stdlib.h> 26#include <string.h> 27#include <math.h> 28 29#include "speak_lib.h" 30#include "speech.h" 31#include "phoneme.h" 32#include "synthesize.h" 33#include "voice.h" 34#include "translate.h" 35 36 37extern FILE *f_log; 38static void SmoothSpect(void); 39 40 41// list of phonemes in a clause 42int n_phoneme_list=0; 43PHONEME_LIST phoneme_list[N_PHONEME_LIST]; 44 45char mbrola_name[20]; 46 47int speed_factor1; 48int speed_factor2; 49int speed_min_sample_len; 50 51static int last_pitch_cmd; 52static int last_amp_cmd; 53static frame_t *last_frame; 54static int last_wcmdq; 55static int pitch_length; 56static int amp_length; 57static int modn_flags; 58 59static int syllable_start; 60static int syllable_end; 61static int syllable_centre; 62 63static voice_t *new_voice=NULL; 64 65int n_soundicon_tab=N_SOUNDICON_SLOTS; 66SOUND_ICON soundicon_tab[N_SOUNDICON_TAB]; 67 68#define RMS_GLOTTAL1 35 // vowel before glottal stop 69#define RMS_START 28 // 14 - 30 70 71#define VOWEL_FRONT_LENGTH 50 72 73 74 75// a dummy phoneme_list entry which looks like a pause 76static PHONEME_LIST next_pause; 77 78 79const char *WordToString(unsigned int word) 80{//======================================== 81// Convert a phoneme mnemonic word into a string 82 int ix; 83 static char buf[5]; 84 85 for(ix=0; ix<3; ix++) 86 buf[ix] = word >> (ix*8); 87 buf[4] = 0; 88 return(buf); 89} 90 91 92void SynthesizeInit() 93{//================== 94 last_pitch_cmd = 0; 95 last_amp_cmd = 0; 96 last_frame = NULL; 97 syllable_centre = -1; 98 99 // initialise next_pause, a dummy phoneme_list entry 100// next_pause.ph = phoneme_tab[phonPAUSE]; // this must be done after voice selection 101 next_pause.type = phPAUSE; 102 next_pause.newword = 0; 103} 104 105 106 107static void EndAmplitude(void) 108{//=========================== 109 if(amp_length > 0) 110 { 111 if(wcmdq[last_amp_cmd][1] == 0) 112 wcmdq[last_amp_cmd][1] = amp_length; 113 amp_length = 0; 114 } 115} 116 117 118 119static void EndPitch(int voice_break) 120{//================================== 121 // posssible end of pitch envelope, fill in the length 122 if((pitch_length > 0) && (last_pitch_cmd >= 0)) 123 { 124 if(wcmdq[last_pitch_cmd][1] == 0) 125 wcmdq[last_pitch_cmd][1] = pitch_length; 126 pitch_length = 0; 127 } 128 129 if(voice_break) 130 { 131 last_wcmdq = -1; 132 last_frame = NULL; 133 syllable_end = wcmdq_tail; 134 SmoothSpect(); 135 syllable_centre = -1; 136 memset(vowel_transition,0,sizeof(vowel_transition)); 137 } 138} // end of Synthesize::EndPitch 139 140 141 142static void DoAmplitude(int amp, unsigned char *amp_env) 143{//===================================================== 144 long *q; 145 146 last_amp_cmd = wcmdq_tail; 147 amp_length = 0; // total length of vowel with this amplitude envelope 148 149 q = wcmdq[wcmdq_tail]; 150 q[0] = WCMD_AMPLITUDE; 151 q[1] = 0; // fill in later from amp_length 152 q[2] = (long)amp_env; 153 q[3] = amp; 154 WcmdqInc(); 155} // end of Synthesize::DoAmplitude 156 157 158 159static void DoPitch(unsigned char *env, int pitch1, int pitch2) 160{//============================================================ 161 long *q; 162 163 EndPitch(0); 164 165 if(pitch1 == 1024) 166 { 167 // pitch was not set 168 pitch1 = 24; 169 pitch2 = 33; 170 env = envelope_data[PITCHfall]; 171 } 172 last_pitch_cmd = wcmdq_tail; 173 pitch_length = 0; // total length of spect with this pitch envelope 174 175 if(pitch2 < 0) 176 pitch2 = 0; 177 178 q = wcmdq[wcmdq_tail]; 179 q[0] = WCMD_PITCH; 180 q[1] = 0; // length, fill in later from pitch_length 181 q[2] = (long)env; 182 q[3] = (pitch1 << 16) + pitch2; 183 WcmdqInc(); 184} // end of Synthesize::DoPitch 185 186 187 188int PauseLength(int pause, int control) 189{//==================================== 190 int len; 191 192 if(control == 0) 193 len = (pause * speed_factor1)/256; 194 else 195 len = (pause * speed_factor2)/256; 196 197 if(len < 5) len = 5; // mS, limit the amount to which pauses can be shortened 198 return(len); 199} 200 201 202static void DoPause(int length, int control) 203{//========================================= 204// control = 1, less shortening at fast speeds 205 int len; 206 207 len = PauseLength(length, control); 208 209 len = (len * samplerate) / 1000; // convert from mS to number of samples 210 211 EndPitch(1); 212 wcmdq[wcmdq_tail][0] = WCMD_PAUSE; 213 wcmdq[wcmdq_tail][1] = len; 214 WcmdqInc(); 215 last_frame = NULL; 216} // end of Synthesize::DoPause 217 218 219extern int seq_len_adjust; // temporary fix to advance the start point for playing the wav sample 220 221 222static int DoSample2(int index, int which, int length_mod, int amp) 223{//================================================================ 224 int length; 225 int length1; 226 int format; 227 int min_length; 228 int start=0; 229 long *q; 230 unsigned char *p; 231 232 index = index & 0x7fffff; 233 p = &wavefile_data[index]; 234 format = p[2]; 235 length1 = (p[1] * 256); 236 length1 += p[0]; // length in bytes 237 238 if(seq_len_adjust > 0) 239 { 240 start = (seq_len_adjust * samplerate)/1000; 241 if(format == 0) 242 start *= 2; 243 length1 -= start; 244 index += start; 245 } 246 247 248 if(length_mod > 0) 249 length = (length1 * length_mod) / 256; 250 else 251 length = length1; 252 253 254 length = (length * speed_factor2)/256; 255 min_length = speed_min_sample_len; 256 if(format==0) 257 min_length *= 2; 258 259 if(length < min_length) 260 length = min_length; 261 262 if(length > length1) 263 length = length1; // don't exceed wavefile length 264 265 if(format==0) 266 length /= 2; // 2 byte samples 267 268 269 index += 4; 270 271 if(amp >= 0) 272 { 273 last_wcmdq = wcmdq_tail; 274 q = wcmdq[wcmdq_tail]; 275 if(which & 0x100) 276 q[0] = WCMD_WAVE2; // mix this with synthesised wave 277 else 278 q[0] = WCMD_WAVE; 279 q[1] = length; // length in samples 280 q[2] = long(&wavefile_data[index]); 281 q[3] = format + (amp << 8); 282 WcmdqInc(); 283 } 284 return(length); 285} // end of Synthesize::DoSample2 286 287 288int DoSample(PHONEME_TAB *ph1, PHONEME_TAB *ph2, int which, int length_mod, int amp) 289{//====================== ========================================================== 290 int index; 291 int match_level; 292 int amp2; 293 int result; 294 295 EndPitch(1); 296 index = LookupSound(ph1,ph2,which & 0xff,&match_level,0); 297 if((index & 0x800000) == 0) 298 return(0); // not wavefile data 299 300 amp2 = wavefile_amp; 301 if(amp != 0) 302 amp2 = (amp * wavefile_amp)/20; 303 304 if(amp == -1) 305 amp2 = amp; 306 307 result = DoSample2(index,which,length_mod,amp2); 308 last_frame = NULL; 309 return(result); 310} // end of Synthesize::DoSample 311 312 313 314 315static frame_t *AllocFrame() 316{//========================= 317 // Allocate a temporary spectrum frame for the wavegen queue. Use a pool which is big 318 // enough to use a round-robin without checks. 319 // Only needed for modifying spectra for blending to consonants 320 321#define N_FRAME_POOL N_WCMDQ 322 static int ix=0; 323 static frame_t frame_pool[N_FRAME_POOL]; 324 325 ix++; 326 if(ix >= N_FRAME_POOL) 327 ix = 0; 328 return(&frame_pool[ix]); 329} 330 331 332static void set_frame_rms(frame_t *fr, int new_rms) 333{//================================================= 334// Each frame includes its RMS amplitude value, so to set a new 335// RMS just adjust the formant amplitudes by the appropriate ratio 336 337 int x; 338 int h; 339 int ix; 340 341 static const short sqrt_tab[200] = { 342 0, 64, 90,110,128,143,156,169,181,192,202,212,221,230,239,247, 343 256,263,271,278,286,293,300,306,313,320,326,332,338,344,350,356, 344 362,367,373,378,384,389,394,399,404,409,414,419,424,429,434,438, 345 443,448,452,457,461,465,470,474,478,483,487,491,495,499,503,507, 346 512,515,519,523,527,531,535,539,543,546,550,554,557,561,565,568, 347 572,576,579,583,586,590,593,596,600,603,607,610,613,617,620,623, 348 627,630,633,636,640,643,646,649,652,655,658,662,665,668,671,674, 349 677,680,683,686,689,692,695,698,701,704,706,709,712,715,718,721, 350 724,726,729,732,735,738,740,743,746,749,751,754,757,759,762,765, 351 768,770,773,775,778,781,783,786,789,791,794,796,799,801,804,807, 352 809,812,814,817,819,822,824,827,829,832,834,836,839,841,844,846, 353 849,851,853,856,858,861,863,865,868,870,872,875,877,879,882,884, 354 886,889,891,893,896,898,900,902}; 355 356 if(fr->rms == 0) return; // check for divide by zero 357 x = (new_rms * 64)/fr->rms; 358 if(x >= 200) x = 199; 359 360 x = sqrt_tab[x]; // sqrt(new_rms/fr->rms)*0x200; 361 362 for(ix=0; ix<N_PEAKS; ix++) 363 { 364 h = fr->fheight[ix] * x; 365 fr->fheight[ix] = h/0x200; 366 } 367} /* end of set_frame_rms */ 368 369 370 371static void formants_reduce_hf(frame_t *fr, int level) 372{//==================================================== 373// change height of peaks 2 to 8, percentage 374 int ix; 375 int x; 376 377 for(ix=2; ix<N_PEAKS; ix++) 378 { 379 x = fr->fheight[ix] * level; 380 fr->fheight[ix] = x/100; 381 } 382} 383 384 385static frame_t *CopyFrame(frame_t *frame1, int copy) 386{//================================================= 387// create a copy of the specified frame in temporary buffer 388 frame_t *frame2; 389 390 if((copy==0) && (frame1->frflags & FRFLAG_COPIED)) 391 { 392 // this frame has already been copied in temporary rw memory 393 return(frame1); 394 } 395 396 frame2 = AllocFrame(); 397 if(frame2 != NULL) 398 { 399 memcpy(frame2,frame1,sizeof(frame_t)); 400 frame2->length = 0; 401 frame2->frflags |= FRFLAG_COPIED; 402 } 403 return(frame2); 404} 405 406 407static frame_t *DuplicateLastFrame(frameref_t *seq, int n_frames, int length) 408{//========================================================================== 409 frame_t *fr; 410 411 seq[n_frames-1].length = length; 412 fr = CopyFrame(seq[n_frames-1].frame,1); 413 seq[n_frames].frame = fr; 414 seq[n_frames].length = 0; 415 return fr; 416} 417 418 419static void AdjustFormants(frame_t *fr, int target, int min, int max, int f1_adj, int f3_adj, int hf_reduce, int flags) 420{//==================================================================================================================== 421 int x; 422 423//hf_reduce = 70; // ?? using fixed amount rather than the parameter?? 424 425 target = (target * voice->formant_factor)/256; 426 427 x = (target - fr->ffreq[2]) / 2; 428 if(x > max) x = max; 429 if(x < min) x = min; 430 fr->ffreq[2] += x; 431 fr->ffreq[3] += f3_adj; 432 433 if(flags & 0x20) 434 { 435 f3_adj = -f3_adj; //. reverse direction for f4,f5 change 436 } 437 fr->ffreq[4] += f3_adj; 438 fr->ffreq[5] += f3_adj; 439 440 if(f1_adj==1) 441 { 442 x = (235 - fr->ffreq[1]); 443 if(x < -100) x = -100; 444 if(x > -60) x = -60; 445 fr->ffreq[1] += x; 446 } 447 if(f1_adj==2) 448 { 449 x = (235 - fr->ffreq[1]); 450 if(x < -300) x = -300; 451 if(x > -150) x = -150; 452 fr->ffreq[1] += x; 453 fr->ffreq[0] += x; 454 } 455 if(f1_adj==3) 456 { 457 x = (100 - fr->ffreq[1]); 458 if(x < -400) x = -400; 459 if(x > -300) x = -400; 460 fr->ffreq[1] += x; 461 fr->ffreq[0] += x; 462 } 463 formants_reduce_hf(fr,hf_reduce); 464} 465 466 467int VowelCloseness(frame_t *fr) 468{//============================ 469// return a value 0-3 depending on the vowel's f1 470 int f1; 471 472 if((f1 = fr->ffreq[1]) < 300) 473 return(3); 474 if(f1 < 400) 475 return(2); 476 if(f1 < 500) 477 return(1); 478 return(0); 479} 480 481 482int FormantTransition2(frameref_t *seq, int &n_frames, unsigned int data1, unsigned int data2, PHONEME_TAB *other_ph, int which) 483{//============================================================================================================================== 484 int ix; 485 int formant; 486 int next_rms; 487 488 int len; 489 int rms; 490 int f1; 491 int f2; 492 int f2_min; 493 int f2_max; 494 int f3_adj; 495 int f3_amp; 496 int flags; 497 int vcolour; 498 499#define N_VCOLOUR 2 500// percentage change for each formant in 256ths 501static short vcolouring[N_VCOLOUR][5] = { 502 {243,272,256,256,256}, // palatal consonant follows 503 {256,256,240,240,240}, // retroflex 504}; 505 506 frame_t *fr = NULL; 507 508 if(n_frames < 2) 509 return(0); 510 511 len = (data1 & 0x3f) * 2; 512 rms = (data1 >> 6) & 0x3f; 513 flags = (data1 >> 12); 514 515 f2 = (data2 & 0x3f) * 50; 516 f2_min = (((data2 >> 6) & 0x1f) - 15) * 50; 517 f2_max = (((data2 >> 11) & 0x1f) - 15) * 50; 518 f3_adj = (((data2 >> 16) & 0x1f) - 15) * 50; 519 f3_amp = ((data2 >> 21) & 0x1f) * 8; 520 f1 = ((data2 >> 26) & 0x7); 521 vcolour = (data2 >> 29); 522 523// fprintf(stderr,"FMT%d %3s %3d-%3d f1=%d f2=%4d %4d %4d f3=%4d %3d\n", 524// which,WordToString(other_ph->mnemonic),len,rms,f1,f2,f2_min,f2_max,f3_adj,f3_amp); 525 526 if(other_ph->mnemonic == '?') 527 flags |= 8; 528 529 if(which == 1) 530 { 531 /* entry to vowel */ 532 fr = CopyFrame(seq[0].frame,0); 533 seq[0].frame = fr; 534 seq[0].length = VOWEL_FRONT_LENGTH; 535 if(len > 0) 536 seq[0].length = len; 537 seq[0].frflags |= FRFLAG_LEN_MOD; // reduce length modification 538 fr->frflags |= FRFLAG_LEN_MOD; 539 540 next_rms = seq[1].frame->rms; 541 542 if(f2 != 0) 543 { 544 if(rms & 0x20) 545 { 546 set_frame_rms(fr,(next_rms * (rms & 0x1f))/30); 547 } 548 AdjustFormants(fr, f2, f2_min, f2_max, f1, f3_adj, f3_amp, flags); 549 550 if((rms & 0x20) == 0) 551 { 552 set_frame_rms(fr,rms*2); 553 } 554 } 555 else 556 { 557 if(flags & 8) 558 set_frame_rms(fr,(next_rms*24)/32); 559 else 560 set_frame_rms(fr,RMS_START); 561 } 562 563 if(flags & 8) 564 { 565// set_frame_rms(fr,next_rms - 5); 566 modn_flags = 0x800 + (VowelCloseness(fr) << 8); 567 } 568 } 569 else 570 { 571 // exit from vowel 572 rms = rms*2; 573 if((f2 != 0) || (flags != 0)) 574 { 575 576 if(flags & 8) 577 { 578 fr = CopyFrame(seq[n_frames-1].frame,0); 579 seq[n_frames-1].frame = fr; 580 rms = RMS_GLOTTAL1; 581 582 // degree of glottal-stop effect depends on closeness of vowel (indicated by f1 freq) 583 modn_flags = 0x400 + (VowelCloseness(fr) << 8); 584 } 585 else 586 { 587 fr = DuplicateLastFrame(seq,n_frames++,len); 588 if(len > 36) 589 seq_len_adjust += (len - 36); 590 591 if(f2 != 0) 592 { 593 AdjustFormants(fr, f2, f2_min, f2_max, f1, f3_adj, f3_amp, flags); 594 } 595 } 596 597 set_frame_rms(fr,rms); 598 599 if((vcolour > 0) && (vcolour <= N_VCOLOUR)) 600 { 601 for(ix=0; ix<n_frames; ix++) 602 { 603 fr = CopyFrame(seq[ix].frame,0); 604 seq[ix].frame = fr; 605 606 for(formant=1; formant<=5; formant++) 607 { 608 int x; 609 x = fr->ffreq[formant] * vcolouring[vcolour-1][formant-1]; 610 fr->ffreq[formant] = x / 256; 611 } 612 } 613 } 614 } 615 } 616 617 if(fr != NULL) 618 { 619 if(flags & 4) 620 fr->frflags |= FRFLAG_FORMANT_RATE; 621 if(flags & 2) 622 fr->frflags |= FRFLAG_BREAK; // don't merge with next frame 623 } 624 625 if(flags & 0x40) 626 DoPause(12,0); // add a short pause after the consonant 627 628 if(flags & 16) 629 return(len); 630 return(0); 631} // end of FormantTransition2 632 633 634 635static void SmoothSpect(void) 636{//========================== 637 // Limit the rate of frequence change of formants, to reduce chirping 638 639 long *q; 640 frame_t *frame; 641 frame_t *frame2; 642 frame_t *frame1; 643 frame_t *frame_centre; 644 int ix; 645 int len; 646 int pk; 647 int modified; 648 int allowed; 649 int diff; 650 651 if(syllable_start == syllable_end) 652 return; 653 654 if((syllable_centre < 0) || (syllable_centre == syllable_start)) 655 { 656 syllable_start = syllable_end; 657 return; 658 } 659 660 q = wcmdq[syllable_centre]; 661 frame_centre = (frame_t *)q[2]; 662 663 // backwards 664 ix = syllable_centre -1; 665 frame = frame2 = frame_centre; 666 for(;;) 667 { 668 if(ix < 0) ix = N_WCMDQ-1; 669 q = wcmdq[ix]; 670 671 if(q[0] == WCMD_PAUSE || q[0] == WCMD_WAVE) 672 break; 673 674 if(q[0] == WCMD_SPECT || q[0] == WCMD_SPECT2) 675 { 676 len = q[1] & 0xffff; 677 678 frame1 = (frame_t *)q[3]; 679 if(frame1 == frame) 680 { 681 q[3] = (long)frame2; 682 frame1 = frame2; 683 } 684 else 685 break; // doesn't follow on from previous frame 686 687 frame = frame2 = (frame_t *)q[2]; 688 modified = 0; 689 690 if(frame->frflags & FRFLAG_BREAK) 691 break; 692 693 if(frame->frflags & FRFLAG_FORMANT_RATE) 694 len = (len * 12)/10; // allow slightly greater rate of change for this frame (was 12/10) 695 696 for(pk=0; pk<6; pk++) 697 { 698 int f1, f2; 699 700 if((frame->frflags & FRFLAG_BREAK_LF) && (pk < 3)) 701 continue; 702 703 f1 = frame1->ffreq[pk]; 704 f2 = frame->ffreq[pk]; 705 706 // backwards 707 if((diff = f2 - f1) > 0) 708 { 709 allowed = f1*2 + f2; 710 } 711 else 712 { 713 allowed = f1 + f2*2; 714 } 715 716 // the allowed change is specified as percentage (%*10) of the frequency 717 // take "frequency" as 1/3 from the lower freq 718 allowed = (allowed * formant_rate[pk])/3000; 719 allowed = (allowed * len)/256; 720 721 if(diff > allowed) 722 { 723 if(modified == 0) 724 { 725 frame2 = CopyFrame(frame,0); 726 modified = 1; 727 } 728 frame2->ffreq[pk] = frame1->ffreq[pk] + allowed; 729 q[2] = (long)frame2; 730 } 731 else 732 if(diff < -allowed) 733 { 734 if(modified == 0) 735 { 736 frame2 = CopyFrame(frame,0); 737 modified = 1; 738 } 739 frame2->ffreq[pk] = frame1->ffreq[pk] - allowed; 740 q[2] = (long)frame2; 741 } 742 } 743 } 744 745 if(ix == syllable_start) 746 break; 747 ix--; 748 } 749 750 // forwards 751 ix = syllable_centre; 752 753 frame = NULL; 754 for(;;) 755 { 756 q = wcmdq[ix]; 757 758 if(q[0] == WCMD_PAUSE || q[0] == WCMD_WAVE) 759 break; 760 761 if(q[0] == WCMD_SPECT || q[0] == WCMD_SPECT2) 762 { 763 764 len = q[1] & 0xffff; 765 766 frame1 = (frame_t *)q[2]; 767 if(frame != NULL) 768 { 769 if(frame1 == frame) 770 { 771 q[2] = (long)frame2; 772 frame1 = frame2; 773 } 774 else 775 break; // doesn't follow on from previous frame 776 } 777 778 frame = frame2 = (frame_t *)q[3]; 779 modified = 0; 780 781 if(frame1->frflags & FRFLAG_BREAK) 782 break; 783 784 if(frame1->frflags & FRFLAG_FORMANT_RATE) 785 len = (len *6)/5; // allow slightly greater rate of change for this frame 786 787 for(pk=0; pk<6; pk++) 788 { 789 int f1, f2; 790 f1 = frame1->ffreq[pk]; 791 f2 = frame->ffreq[pk]; 792 793 // forwards 794 if((diff = f2 - f1) > 0) 795 { 796 allowed = f1*2 + f2; 797 } 798 else 799 { 800 allowed = f1 + f2*2; 801 } 802 allowed = (allowed * formant_rate[pk])/3000; 803 allowed = (allowed * len)/256; 804 805 if(diff > allowed) 806 { 807 if(modified == 0) 808 { 809 frame2 = CopyFrame(frame,0); 810 modified = 1; 811 } 812 frame2->ffreq[pk] = frame1->ffreq[pk] + allowed; 813 q[3] = (long)frame2; 814 } 815 else 816 if(diff < -allowed) 817 { 818 if(modified == 0) 819 { 820 frame2 = CopyFrame(frame,0); 821 modified = 1; 822 } 823 frame2->ffreq[pk] = frame1->ffreq[pk] - allowed; 824 q[3] = (long)frame2; 825 } 826 } 827 } 828 829 ix++; 830 if(ix >= N_WCMDQ) ix = 0; 831 if(ix == syllable_end) 832 break; 833 } 834 835 syllable_start = syllable_end; 836} 837 838 839static void StartSyllable(void) 840{//============================ 841 // start of syllable, if not already started 842 if(syllable_end == syllable_start) 843 syllable_end = wcmdq_tail; 844} 845 846 847int DoSpect(PHONEME_TAB *this_ph, PHONEME_TAB *prev_ph, PHONEME_TAB *next_ph, 848 int which, PHONEME_LIST *plist, int modulation) 849{//=================================================================================== 850 // which 1 start of phoneme, 2 body and end 851 // length_mod: 256 = 100% 852 // modulation: -1 = don't write to wcmdq 853 854 int n_frames; 855 frameref_t *frames; 856 int frameix; 857 frame_t *frame1; 858 frame_t *frame2; 859 frame_t *fr; 860 int ix; 861 long *q; 862 int len; 863 int match_level; 864 int frame_length; 865 int frame1_length; 866 int frame2_length; 867 int length_factor; 868 int length_mod; 869 int total_len = 0; 870 static int wave_flag = 0; 871 int wcmd_spect = WCMD_SPECT; 872 873 length_mod = plist->length; 874 if(length_mod==0) length_mod=256; 875 876if(which==1) 877{ 878 // limit the shortening of sonorants before shortened (eg. unstressed vowels) 879 if((this_ph->type==phLIQUID) || (prev_ph->type==phLIQUID) || (prev_ph->type==phNASAL)) 880 { 881 if(length_mod < (len = translator->langopts.param[LOPT_SONORANT_MIN])) 882 { 883 length_mod = len; 884 } 885 } 886} 887 888 modn_flags = 0; 889 frames = LookupSpect(this_ph,prev_ph,next_ph,which,&match_level,&n_frames, plist); 890 if(frames == NULL) 891 return(0); // not found 892 893 if(wavefile_ix == 0) 894 { 895 if(wave_flag) 896 { 897 // cancel any wavefile that was playing previously 898 wcmd_spect = WCMD_SPECT2; 899 wave_flag = 0; 900 } 901 else 902 { 903 wcmd_spect = WCMD_SPECT; 904 } 905 } 906 907 frame1 = frames[0].frame; 908 frame1_length = frames[0].length; 909 if(last_frame != NULL) 910 { 911 if(((last_frame->length < 2) || (last_frame->frflags & FRFLAG_VOWEL_CENTRE)) 912 && !(last_frame->frflags & FRFLAG_BREAK)) 913 { 914 // last frame of previous sequence was zero-length, replace with first of this sequence 915 wcmdq[last_wcmdq][3] = (long)frame1; 916 917 if(last_frame->frflags & FRFLAG_BREAK_LF) 918 { 919 // but flag indicates keep HF peaks in last segment 920 fr = CopyFrame(frame1,1); 921 for(ix=3; ix<N_PEAKS; ix++) 922 { 923 fr->ffreq[ix] = last_frame->ffreq[ix]; 924 fr->fheight[ix] = last_frame->fheight[ix]; 925 } 926 wcmdq[last_wcmdq][3] = (long)fr; 927 } 928 } 929 } 930 931 if((this_ph->type == phVOWEL) && (which == 2)) 932 { 933 SmoothSpect(); // process previous syllable 934 935 // remember the point in the output queue of the centre of the vowel 936 syllable_centre = wcmdq_tail; 937 } 938 939 frame_length = frame1_length; 940 for(frameix=1; frameix<n_frames; frameix++) 941 { 942 frame2 = frames[frameix].frame; 943 frame2_length = frames[frameix].length; 944 945 if((wavefile_ix != 0) && ((frame1->frflags & FRFLAG_DEFER_WAV)==0)) 946 { 947 // there is a wave file to play along with this synthesis 948 seq_len_adjust = 0; 949 DoSample2(wavefile_ix,which+0x100,0,wavefile_amp); 950 wave_flag = 1; 951 wavefile_ix = 0; 952 } 953 954 length_factor = length_mod; 955 if(frame1->frflags & FRFLAG_LEN_MOD) // reduce effect of length mod 956 { 957 length_factor = (length_mod*4 + 256*3)/7; 958 } 959 len = (frame_length * samplerate)/1000; 960 len = (len * length_factor)/256; 961 962 if(modulation >= 0) 963 { 964 if(frame1->frflags & FRFLAG_MODULATE) 965 { 966 modulation = 6; 967 } 968 if((frameix == n_frames-1) && (modn_flags & 0xf00)) 969 modulation |= modn_flags; // before or after a glottal stop 970 } 971 972 pitch_length += len; 973 amp_length += len; 974 975 if(frame_length < 2) 976 { 977 last_frame = NULL; 978 frame_length = frame2_length; 979 frame1 = frame2; 980 } 981 else 982 { 983 last_wcmdq = wcmdq_tail; 984 985 if(modulation >= 0) 986 { 987 q = wcmdq[wcmdq_tail]; 988 q[0] = wcmd_spect; 989 q[1] = len + (modulation << 16); 990 q[2] = long(frame1); 991 q[3] = long(frame2); 992 993 WcmdqInc(); 994 } 995 last_frame = frame1 = frame2; 996 frame_length = frame2_length; 997 total_len += len; 998 } 999 } 1000 return(total_len); 1001} // end of Synthesize::DoSpect 1002 1003 1004static void DoMarker(int type, int char_posn, int length, int value) 1005{//================================================================= 1006// This could be used to return an index to the word currently being spoken 1007// Type 1=word, 2=sentence, 3=named marker, 4=play audio, 5=end 1008 wcmdq[wcmdq_tail][0] = WCMD_MARKER; 1009 wcmdq[wcmdq_tail][1] = type; 1010 wcmdq[wcmdq_tail][2] = (char_posn & 0xffffff) | (length << 24); 1011 wcmdq[wcmdq_tail][3] = value; 1012 WcmdqInc(); 1013 1014} // end of Synthesize::DoMarker 1015 1016 1017void DoVoiceChange(voice_t *v) 1018{//=========================== 1019// allocate memory for a copy of the voice data, and free it in wavegenfill() 1020 voice_t *v2; 1021 1022 v2 = (voice_t *)malloc(sizeof(voice_t)); 1023 memcpy(v2,v,sizeof(voice_t)); 1024 wcmdq[wcmdq_tail][0] = WCMD_VOICE; 1025 wcmdq[wcmdq_tail][1] = (long)(v2); 1026 WcmdqInc(); 1027} 1028 1029 1030static void DoEmbedded(int &embix, int sourceix) 1031{//============================================= 1032 // There were embedded commands in the text at this point 1033 unsigned int word; // bit 7=last command for this word, bits 5,6 sign, bits 0-4 command 1034 unsigned int value; 1035 int command; 1036 1037 do { 1038 word = embedded_list[embix++]; 1039 value = word >> 8; 1040 command = word & 0x7f; 1041 1042 switch(command & 0x1f) 1043 { 1044 case EMBED_S: // speed 1045 SetEmbedded((command & 0x60) + EMBED_S2,value); // adjusts embedded_value[EMBED_S2] 1046 SetSpeed(2); 1047 break; 1048 1049 case EMBED_I: // play dynamically loaded wav data (sound icon) 1050 if((int)value < n_soundicon_tab) 1051 { 1052 if(soundicon_tab[value].length != 0) 1053 { 1054 DoPause(10,0); // ensure a break in the speech 1055 wcmdq[wcmdq_tail][0] = WCMD_WAVE; 1056 wcmdq[wcmdq_tail][1] = soundicon_tab[value].length; 1057 wcmdq[wcmdq_tail][2] = (long)soundicon_tab[value].data + 44; // skip WAV header 1058 wcmdq[wcmdq_tail][3] = 0x1500; // 16 bit data, amp=21 1059 WcmdqInc(); 1060 } 1061 } 1062 break; 1063 1064 case EMBED_M: // named marker 1065 DoMarker(espeakEVENT_MARK, (sourceix & 0x7ff) + clause_start_char, 0, value); 1066 break; 1067 1068 case EMBED_U: // play sound 1069 DoMarker(espeakEVENT_PLAY, count_characters+1, 0, value); // always occurs at end of clause 1070 break; 1071 1072 default: 1073 DoPause(10,0); // ensure a break in the speech 1074 wcmdq[wcmdq_tail][0] = WCMD_EMBEDDED; 1075 wcmdq[wcmdq_tail][1] = command; 1076 wcmdq[wcmdq_tail][2] = value; 1077 WcmdqInc(); 1078 break; 1079 } 1080 } while ((word & 0x80) == 0); 1081} 1082 1083 1084void SwitchDictionary() 1085{//==================== 1086} 1087 1088 1089int Generate(PHONEME_LIST *phoneme_list, int *n_ph, int resume) 1090{//============================================================ 1091 static int ix; 1092 static int embedded_ix; 1093 static int word_count; 1094 PHONEME_LIST *prev; 1095 PHONEME_LIST *next; 1096 PHONEME_LIST *next2; 1097 PHONEME_LIST *p; 1098 int released; 1099 int stress; 1100 int modulation; 1101 int pre_voiced; 1102 int free_min; 1103 unsigned char *pitch_env=NULL; 1104 unsigned char *amp_env; 1105 PHONEME_TAB *ph; 1106 PHONEME_TAB *prev_ph; 1107 static int sourceix=0; 1108 1109#ifdef TEST_MBROLA 1110 if(mbrola_name[0] != 0) 1111 return(MbrolaGenerate(phoneme_list,n_ph,resume)); 1112#endif 1113 1114 if(option_quiet) 1115 return(0); 1116 1117 if(resume == 0) 1118 { 1119 ix = 1; 1120 embedded_ix=0; 1121 word_count = 0; 1122 pitch_length = 0; 1123 amp_length = 0; 1124 last_frame = NULL; 1125 last_wcmdq = -1; 1126 syllable_start = wcmdq_tail; 1127 syllable_end = wcmdq_tail; 1128 syllable_centre = -1; 1129 last_pitch_cmd = -1; 1130 memset(vowel_transition,0,sizeof(vowel_transition)); 1131 } 1132 1133 while(ix < (*n_ph)) 1134 { 1135 p = &phoneme_list[ix]; 1136 1137 if(p->type == phPAUSE) 1138 free_min = 5; 1139 else 1140 if(p->type != phVOWEL) 1141 free_min = 10; // we need less Q space for non-vowels, and we need to generate phonemes after a vowel so that the pitch_length is filled in 1142 else 1143 free_min = MIN_WCMDQ; // 22 1144 1145 if(WcmdqFree() <= free_min) 1146 return(1); // wait 1147 1148 prev = &phoneme_list[ix-1]; 1149 next = &phoneme_list[ix+1]; 1150 next2 = &phoneme_list[ix+2]; 1151 1152 if(p->synthflags & SFLAG_EMBEDDED) 1153 { 1154 DoEmbedded(embedded_ix, p->sourceix); 1155 } 1156 1157 if(p->newword) 1158 { 1159 if(translator->langopts.param[LOPT_WORD_MERGE] == 0) 1160 last_frame = NULL; 1161 1162 sourceix = (p->sourceix & 0x7ff) + clause_start_char; 1163 1164 if(p->newword & 4) 1165 DoMarker(espeakEVENT_SENTENCE, sourceix, 0, count_sentences); // start of sentence 1166 1167// if(p->newword & 2) 1168// DoMarker(espeakEVENT_END, count_characters, 0, count_sentences); // end of clause 1169 1170 if(p->newword & 1) 1171 DoMarker(espeakEVENT_WORD, sourceix, p->sourceix >> 11, clause_start_word + word_count++); 1172 } 1173 1174 EndAmplitude(); 1175 1176 if(p->prepause > 0) 1177 DoPause(p->prepause,1); 1178 1179 if(option_phoneme_events && (p->type != phVOWEL)) 1180 { 1181 // Note, for vowels, do the phoneme event after the vowel-start 1182 DoMarker(espeakEVENT_PHONEME, sourceix, 0, p->ph->mnemonic); 1183 } 1184 1185 switch(p->type) 1186 { 1187 case phPAUSE: 1188 DoPause(p->length,0); 1189 break; 1190 1191 case phSTOP: 1192 released = 0; 1193 if(next->type==phVOWEL) released = 1; 1194 if(next->type==phLIQUID && !next->newword) released = 1; 1195 1196 if(released) 1197 DoSample(p->ph,next->ph,2,0,0); 1198 else 1199 DoSample(p->ph,phoneme_tab[phonPAUSE],2,0,0); 1200 break; 1201 1202 case phFRICATIVE: 1203 if(p->synthflags & SFLAG_LENGTHEN) 1204 DoSample(p->ph,next->ph,2,p->length,0); // play it twice for [s:] etc. 1205 DoSample(p->ph,next->ph,2,p->length,0); 1206 break; 1207 1208 case phVSTOP: 1209 pre_voiced = 0; 1210 if(next->type==phVOWEL) 1211 { 1212 DoAmplitude(p->amp,NULL); 1213 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1214 pre_voiced = 1; 1215 } 1216 else 1217 if((next->type==phLIQUID) && !next->newword) 1218 { 1219 DoAmplitude(next->amp,NULL); 1220 DoPitch(envelope_data[next->env],next->pitch1,next->pitch2); 1221 pre_voiced = 1; 1222 } 1223 else 1224 { 1225 if(last_pitch_cmd < 0) 1226 { 1227 DoAmplitude(next->amp,NULL); 1228 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1229 } 1230 } 1231 1232 if((prev->type==phVOWEL) || (prev->ph->phflags & phVOWEL2)) 1233 { 1234 // a period of voicing before the release 1235 DoSpect(p->ph,phoneme_tab[phonSCHWA],next->ph,1,p,0); 1236 if(p->synthflags & SFLAG_LENGTHEN) 1237 { 1238 DoPause(20,0); 1239 DoSpect(p->ph,phoneme_tab[phonSCHWA],next->ph,1,p,0); 1240 } 1241 } 1242 else 1243 { 1244 if(p->synthflags & SFLAG_LENGTHEN) 1245 { 1246 DoPause(50,0); 1247 } 1248 } 1249 1250 if(pre_voiced) 1251 { 1252 // followed by a vowel, or liquid + vowel 1253 StartSyllable(); 1254 DoSpect(p->ph,prev->ph,next->ph,2,p,0); 1255 } 1256 else 1257 { 1258// if((prev->type != phVOWEL) && ((prev->ph->phflags & phVOICED)==0) && ((next->ph->phflags & phVOICED)==0)) 1259// DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE_SHORT],2,p,0); 1260// else 1261 DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE],2,p,0); 1262// DoSpect(p->ph,prev->ph,next->ph,2,p,0); 1263 } 1264 break; 1265 1266 case phVFRICATIVE: 1267 if(next->type==phVOWEL) 1268 { 1269 DoAmplitude(p->amp,NULL); 1270 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1271 } 1272 else 1273 if(next->type==phLIQUID) 1274 { 1275 DoAmplitude(next->amp,NULL); 1276 DoPitch(envelope_data[next->env],next->pitch1,next->pitch2); 1277 } 1278 else 1279 { 1280 if(last_pitch_cmd < 0) 1281 { 1282 DoAmplitude(p->amp,NULL); 1283 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1284 } 1285 } 1286 1287 if((next->type==phVOWEL) || ((next->type==phLIQUID)) && (next->newword==0)) // ?? test 14.Aug.2007 1288 { 1289 StartSyllable(); 1290 if(p->synthflags & SFLAG_LENGTHEN) 1291 DoSpect(p->ph,prev->ph,next->ph,2,p,0); 1292 DoSpect(p->ph,prev->ph,next->ph,2,p,0); 1293 } 1294 else 1295 { 1296 if(p->synthflags & SFLAG_LENGTHEN) 1297 DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE],2,p,0); 1298 DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE],2,p,0); 1299 } 1300 break; 1301 1302 case phNASAL: 1303 if(!(p->synthflags & SFLAG_SEQCONTINUE)) 1304 { 1305 DoAmplitude(p->amp,NULL); 1306 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1307 } 1308 1309 if(prev->type==phNASAL) 1310 { 1311 last_frame = NULL; 1312 } 1313 1314 if(next->type==phVOWEL) 1315 { 1316 StartSyllable(); 1317 DoSpect(p->ph,prev->ph,next->ph,1,p,0); 1318 } 1319 else 1320 if(prev->type==phVOWEL && (p->synthflags & SFLAG_SEQCONTINUE)) 1321 { 1322 DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE],2,p,0); 1323 } 1324 else 1325 { 1326 last_frame = NULL; // only for nasal ? 1327 if(next->type == phLIQUID) 1328 DoSpect(p->ph,prev->ph,phoneme_tab[phonSONORANT],2,p,0); 1329 else 1330 DoSpect(p->ph,prev->ph,phoneme_tab[phonPAUSE],2,p,0); 1331 last_frame = NULL; 1332 } 1333 1334 break; 1335 1336 case phLIQUID: 1337 modulation = 0; 1338 if(p->ph->phflags & phTRILL) 1339 modulation = 5; 1340 1341 prev_ph = prev->ph; 1342// if(p->newword) 1343// prev_ph = phoneme_tab[phonPAUSE]; // pronounce fully at the start of a word 1344 1345 if(!(p->synthflags & SFLAG_SEQCONTINUE)) 1346 { 1347 DoAmplitude(p->amp,NULL); 1348 DoPitch(envelope_data[p->env],p->pitch1,p->pitch2); 1349 } 1350 1351 if(prev->type==phNASAL) 1352 { 1353 last_frame = NULL; 1354 } 1355 1356 if(next->type==phVOWEL) 1357 { 1358 StartSyllable(); 1359 DoSpect(p->ph,prev_ph,next->ph,1,p,modulation); // (,)r 1360 } 1361 else 1362 if(prev->type==phVOWEL && (p->synthflags & SFLAG_SEQCONTINUE)) 1363 { 1364 DoSpect(p->ph,prev_ph,next->ph,1,p,modulation); 1365 } 1366 else 1367 { 1368 DoSpect(p->ph,prev_ph,next->ph,1,p,modulation); 1369 } 1370 1371 break; 1372 1373 case phVOWEL: 1374 ph = p->ph; 1375 stress = p->tone & 0xf; 1376 1377 // vowel transition from the preceding phoneme 1378 vowel_transition0 = vowel_transition[0]; 1379 vowel_transition1 = vowel_transition[1]; 1380 1381 pitch_env = envelope_data[p->env]; 1382 amp_env = NULL; 1383 if(p->tone_ph != 0) 1384 { 1385 pitch_env = LookupEnvelope(phoneme_tab[p->tone_ph]->spect); 1386 amp_env = LookupEnvelope(phoneme_tab[p->tone_ph]->after); 1387 } 1388 1389 StartSyllable(); 1390 1391 modulation = 2; 1392 if(stress <= 1) 1393 modulation = 1; // 16ths 1394 else 1395 if(stress >= 7) 1396 modulation = 3; 1397 1398 if(prev->type == phVSTOP || prev->type == phVFRICATIVE) 1399 { 1400 DoAmplitude(p->amp,amp_env); 1401 DoPitch(pitch_env,p->pitch1,p->pitch2); // don't use prevocalic rising tone 1402 DoSpect(ph,prev->ph,next->ph,1,p,modulation); 1403 } 1404 else 1405 if(prev->type==phLIQUID || prev->type==phNASAL) 1406 { 1407 DoAmplitude(p->amp,amp_env); 1408 DoSpect(ph,prev->ph,next->ph,1,p,modulation); // continue with pre-vocalic rising tone 1409 DoPitch(pitch_env,p->pitch1,p->pitch2); 1410 } 1411 else 1412 { 1413 if(!(p->synthflags & SFLAG_SEQCONTINUE)) 1414 { 1415 DoAmplitude(p->amp,amp_env); 1416 DoPitch(pitch_env,p->pitch1,p->pitch2); 1417 } 1418 1419 DoSpect(ph,prev->ph,next->ph,1,p,modulation); 1420 } 1421 1422 if(option_phoneme_events) 1423 { 1424 DoMarker(espeakEVENT_PHONEME, sourceix, 0, p->ph->mnemonic); 1425 } 1426 1427 DoSpect(p->ph,prev->ph,next->ph,2,p,modulation); 1428 1429 memset(vowel_transition,0,sizeof(vowel_transition)); 1430 break; 1431 } 1432 ix++; 1433 } 1434 EndPitch(1); 1435 if(*n_ph > 0) 1436 { 1437 DoMarker(espeakEVENT_END, count_characters, 0, count_sentences); // end of clause 1438 *n_ph = 0; 1439 } 1440 1441 return(0); // finished the phoneme list 1442} // end of Generate 1443 1444 1445 1446 1447static int timer_on = 0; 1448static int paused = 0; 1449 1450int SynthOnTimer() 1451{//=============== 1452 if(!timer_on) 1453 { 1454 return(WavegenCloseSound()); 1455 } 1456 1457 do { 1458 if(Generate(phoneme_list,&n_phoneme_list,1)==0) 1459 { 1460 SpeakNextClause(NULL,NULL,1); 1461 } 1462 } while(skipping_text); 1463 1464 return(0); 1465} 1466 1467 1468int SynthStatus() 1469{//============== 1470 return(timer_on | paused); 1471} 1472 1473 1474 1475int SpeakNextClause(FILE *f_in, const void *text_in, int control) 1476{//============================================================== 1477// Speak text from file (f_in) or memory (text_in) 1478// control 0: start 1479// either f_in or text_in is set, the other must be NULL 1480 1481// The other calls have f_in and text_in = NULL 1482// control 1: speak next text 1483// 2: stop 1484// 3: pause (toggle) 1485// 4: is file being read (0=no, 1=yes) 1486// 5: interrupt and flush current text. 1487 1488 int clause_tone; 1489 char *voice_change; 1490 static FILE *f_text=NULL; 1491 static const void *p_text=NULL; 1492 1493 if(control == 4) 1494 { 1495 if((f_text == NULL) && (p_text == NULL)) 1496 return(0); 1497 else 1498 return(1); 1499 } 1500 1501 if(control == 2) 1502 { 1503 // stop speaking 1504 timer_on = 0; 1505 p_text = NULL; 1506 if(f_text != NULL) 1507 { 1508 fclose(f_text); 1509 f_text=NULL; 1510 } 1511 n_phoneme_list = 0; 1512 WcmdqStop(); 1513 return(0); 1514 } 1515 1516 if(control == 3) 1517 { 1518 // toggle pause 1519 if(paused == 0) 1520 { 1521 timer_on = 0; 1522 paused = 2; 1523 } 1524 else 1525 { 1526 WavegenOpenSound(); 1527 timer_on = 1; 1528 paused = 0; 1529 Generate(phoneme_list,&n_phoneme_list,0); // re-start from beginning of clause 1530 } 1531 return(0); 1532 } 1533 1534 if(control == 5) 1535 { 1536 // stop speaking, but continue looking for text 1537 n_phoneme_list = 0; 1538 WcmdqStop(); 1539 return(0); 1540 } 1541 1542 if((f_in != NULL) || (text_in != NULL)) 1543 { 1544 f_text = f_in; 1545 p_text = text_in; 1546 timer_on = 1; 1547 paused = 0; 1548 } 1549 1550 if((f_text==NULL) && (p_text==NULL)) 1551 { 1552 skipping_text = 0; 1553 timer_on = 0; 1554 return(0); 1555 } 1556 1557 if((f_text != NULL) && feof(f_text)) 1558 { 1559 timer_on = 0; 1560 fclose(f_text); 1561 f_text=NULL; 1562 return(0); 1563 } 1564 1565 if(current_phoneme_table != voice->phoneme_tab_ix) 1566 { 1567 SelectPhonemeTable(voice->phoneme_tab_ix); 1568 } 1569 1570 // read the next clause from the input text file, translate it, and generate 1571 // entries in the wavegen command queue 1572 p_text = translator->TranslateClause(f_text,p_text,&clause_tone,&voice_change); 1573 1574 translator->CalcPitches(clause_tone); 1575 translator->CalcLengths(); 1576 1577 translator->GetTranslatedPhonemeString(translator->phon_out,sizeof(translator->phon_out)); 1578 if(option_phonemes > 0) 1579 { 1580 fprintf(f_trans,"%s\n",translator->phon_out); 1581 1582 if(!iswalpha(0x010d)) 1583 { 1584 // check that c-caron is recognized as an alphabetic character 1585 fprintf(stderr,"Warning: Accented letters are not recognized, eg: U+010D\nSet LC_CTYPE to a UTF-8 locale\n"); 1586 } 1587 } 1588 if(phoneme_callback != NULL) 1589 { 1590 phoneme_callback(translator->phon_out); 1591 } 1592 1593 1594 if(skipping_text) 1595 { 1596 n_phoneme_list = 0; 1597 return(1); 1598 } 1599 1600 if(mbrola_name[0] != 0) 1601 { 1602#ifdef USE_MBROLA_LIB 1603 MbrolaTranslate(phoneme_list,n_phoneme_list,NULL); 1604#else 1605 MbrolaTranslate(phoneme_list,n_phoneme_list,stdout); 1606#endif 1607 } 1608 1609 Generate(phoneme_list,&n_phoneme_list,0); 1610 WavegenOpenSound(); 1611 1612 if(voice_change != NULL) 1613 { 1614 // voice change at the end of the clause (i.e. clause was terminated by a voice change) 1615 new_voice = LoadVoiceVariant(voice_change,0); // add a Voice instruction to wavegen at the end of the clause 1616 } 1617 1618 if(new_voice) 1619 { 1620 // finished the current clause, now change the voice if there was an embedded 1621 // change voice command at the end of it (i.e. clause was broken at the change voice command) 1622 DoVoiceChange(voice); 1623 new_voice = NULL; 1624 } 1625 1626 return(1); 1627} // end of SpeakNextClause 1628