sound/svox/patches/0003-pico2wave-Convert-text-to-.wav-using-svox-text-to-sp.patch

   1 From 8bec80dccc9f4fe147a500486813f4e89a0d56d8 Mon Sep 17 00:00:00 2001
   2 From: Mathieu Parent <math.parent@gmail.com>
   3 Date: Sun, 25 Oct 2009 15:19:01 +0100
   4 Subject: [PATCH 3/7] pico2wave: Convert text to .wav using svox text-to-speech system.
   5
   6 ---
   7  pico/.gitignore      |    1 +
   8  pico/Makefile.am     |    7 +
   9  pico/bin/pico2wave.c |  341 ++++++++++++++++++++++++++++++++++++++++++++++++++
  10  pico/configure.in    |    3 +
  11  4 files changed, 352 insertions(+), 0 deletions(-)
  12  create mode 100644 pico/bin/pico2wave.c
  13
  14 diff --git a/pico/.gitignore b/pico/.gitignore
  15 index 4235569..a110298 100644
  16 --- a/pico/.gitignore
  17 +++ b/pico/.gitignore
  18 @@ -29,4 +29,5 @@ libtool
  19  *.lo
  20  .libs
  21  libttspico.la
  22 +pico2wave
  23
  24 diff --git a/pico/Makefile.am b/pico/Makefile.am
  25 index 6d8a10c..0d9472d 100644
  26 --- a/pico/Makefile.am
  27 +++ b/pico/Makefile.am
  28 @@ -34,3 +34,10 @@ libttspico_la_SOURCES = \
  29         lib/picotrns.c \
  30         lib/picowa.c
  31
  32 +bin_PROGRAMS = pico2wave
  33 +pico2wave_SOURCES = \
  34 +       bin/pico2wave.c
  35 +pico2wave_LDADD = \
  36 +       libttspico.la -lm -lpopt
  37 +pico2wave_CFLAGS = -Wall -I lib
  38 +
  39 diff --git a/pico/bin/pico2wave.c b/pico/bin/pico2wave.c
  40 new file mode 100644
  41 index 0000000..0c035a7
  42 --- /dev/null
  43 +++ b/pico/bin/pico2wave.c
  44 @@ -0,0 +1,342 @@
  45 +/* pico2wave.c
  46 +
  47 + * Copyright (C) 2009 Mathieu Parent <math.parent@gmail.com>
  48 + *
  49 + * Licensed under the Apache License, Version 2.0 (the "License");
  50 + * you may not use this file except in compliance with the License.
  51 + * You may obtain a copy of the License at
  52 + *
  53 + *     http://www.apache.org/licenses/LICENSE-2.0
  54 + *
  55 + * Unless required by applicable law or agreed to in writing, software
  56 + * distributed under the License is distributed on an "AS IS" BASIS,
  57 + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  58 + * See the License for the specific language governing permissions and
  59 + * limitations under the License.
  60 + *
  61 + *   Convert text to .wav using svox text-to-speech system.
  62 + *
  63 + */
  64 +
  65 +
  66 +#include <popt.h>
  67 +#include <stdio.h>
  68 +#include <stdint.h>
  69 +#include <stdlib.h>
  70 +#include <string.h>
  71 +
  72 +#include <picoapi.h>
  73 +#include <picoapid.h>
  74 +#include <picoos.h>
  75 +
  76 +
  77 +/* adaptation layer defines */
  78 +#define PICO_MEM_SIZE       2500000
  79 +#define DummyLen 100000000
  80 +
  81 +/* string constants */
  82 +#define MAX_OUTBUF_SIZE     128
  83 +const char * PICO_LINGWARE_PATH             = "./lang/";
  84 +const char * PICO_VOICE_NAME                = "PicoVoice";
  85 +
  86 +/* supported voices
  87 +   Pico does not separately specify the voice and locale.   */
  88 +const char * picoSupportedLangIso3[]        = { "eng",              "eng",              "deu",              "spa",              "fra",              "ita" };
  89 +const char * picoSupportedCountryIso3[]     = { "USA",              "GBR",              "DEU",              "ESP",              "FRA",              "ITA" };
  90 +const char * picoSupportedLang[]            = { "en-US",            "en-GB",            "de-DE",            "es-ES",            "fr-FR",            "it-IT" };
  91 +const char * picoInternalLang[]             = { "en-US",            "en-GB",            "de-DE",            "es-ES",            "fr-FR",            "it-IT" };
  92 +const char * picoInternalTaLingware[]       = { "en-US_ta.bin",     "en-GB_ta.bin",     "de-DE_ta.bin",     "es-ES_ta.bin",     "fr-FR_ta.bin",     "it-IT_ta.bin" };
  93 +const char * picoInternalSgLingware[]       = { "en-US_lh0_sg.bin", "en-GB_kh0_sg.bin", "de-DE_gl0_sg.bin", "es-ES_zl0_sg.bin", "fr-FR_nk0_sg.bin", "it-IT_cm0_sg.bin" };
  94 +const char * picoInternalUtppLingware[]     = { "en-US_utpp.bin",   "en-GB_utpp.bin",   "de-DE_utpp.bin",   "es-ES_utpp.bin",   "fr-FR_utpp.bin",   "it-IT_utpp.bin" };
  95 +const int picoNumSupportedVocs              = 6;
  96 +
  97 +/* adapation layer global variables */
  98 +void *          picoMemArea         = NULL;
  99 +pico_System     picoSystem          = NULL;
 100 +pico_Resource   picoTaResource      = NULL;
 101 +pico_Resource   picoSgResource      = NULL;
 102 +pico_Resource   picoUtppResource    = NULL;
 103 +pico_Engine     picoEngine          = NULL;
 104 +pico_Char *     picoTaFileName      = NULL;
 105 +pico_Char *     picoSgFileName      = NULL;
 106 +pico_Char *     picoUtppFileName    = NULL;
 107 +pico_Char *     picoTaResourceName  = NULL;
 108 +pico_Char *     picoSgResourceName  = NULL;
 109 +pico_Char *     picoUtppResourceName = NULL;
 110 +int     picoSynthAbort = 0;
 111 +
 112 +
 113 +int main(int argc, const char *argv[]) {
 114 +    char * wavefile = NULL;
 115 +    char * lang = "en-US";
 116 +    int langIndex = -1, langIndexTmp = -1;
 117 +    char * text;
 118 +    int8_t * buffer;
 119 +    size_t bufferSize = 256;
 120 +
 121 +    /* Parsing options */
 122 +       poptContext optCon; /* context for parsing command-line options */
 123 +       int opt; /* used for argument parsing */
 124 +
 125 +       struct poptOption optionsTable[] = {
 126 +               { "wave", 'w', POPT_ARG_STRING, &wavefile, 0,
 127 +                 "Write output to this WAV file (extension SHOULD be .wav)", "filename.wav" },
 128 +               { "lang", 'l', POPT_ARG_STRING | POPT_ARGFLAG_SHOW_DEFAULT, &lang, 0,
 129 +                 "Language", "lang" },
 130 +               POPT_AUTOHELP
 131 +               POPT_TABLEEND
 132 +       };
 133 +       optCon = poptGetContext(NULL, argc, argv, optionsTable, POPT_CONTEXT_POSIXMEHARDER);
 134 +    poptSetOtherOptionHelp(optCon, "<words>");
 135 +
 136 +    /* Reporting about invalid extra options */
 137 +       while ((opt = poptGetNextOpt(optCon)) != -1) {
 138 +               switch (opt) {
 139 +               default:
 140 +                       fprintf(stderr, "Invalid option %s: %s\n",
 141 +                               poptBadOption(optCon, 0), poptStrerror(opt));
 142 +                       poptPrintHelp(optCon, stderr, 0);
 143 +                       exit(1);
 144 +               }
 145 +       }
 146 +
 147 +    /* Mandatory option: --wave */
 148 +       if(!wavefile) {
 149 +               fprintf(stderr, "Mandatory option: %s\n\n",
 150 +                       "--wave=filename.wav");
 151 +               poptPrintHelp(optCon, stderr, 0);
 152 +               exit(1);
 153 +       }
 154 +       /* option: --lang */
 155 +       for(langIndexTmp =0; langIndexTmp<picoNumSupportedVocs; langIndexTmp++) {
 156 +           if(!strcmp(picoSupportedLang[langIndexTmp], lang)) {
 157 +               langIndex = langIndexTmp;
 158 +               break;
 159 +           }
 160 +       }
 161 +       if(langIndex == -1) {
 162 +               fprintf(stderr, "Unknown language: %s\nValid languages:\n",
 163 +                       lang);
 164 +           for(langIndexTmp =0; langIndexTmp<picoNumSupportedVocs; langIndexTmp++) {
 165 +               fprintf(stderr, "%s\n", picoSupportedLang[langIndexTmp]);
 166 +           }
 167 +           lang = "en-US";
 168 +               fprintf(stderr, "\n");
 169 +               poptPrintHelp(optCon, stderr, 0);
 170 +               exit(1);
 171 +       }
 172 +
 173 +       /* Remaining argument is <words> */
 174 +       const char **extra_argv;
 175 +       extra_argv = poptGetArgs(optCon);
 176 +    if(extra_argv) {
 177 +               text = (char *) &(*extra_argv)[0];
 178 +    } else {
 179 +        //TODO: stdin not supported yet.
 180 +               fprintf(stderr, "Missing argument: %s\n\n",
 181 +                       "<words>");
 182 +               poptPrintHelp(optCon, stderr, 0);
 183 +               exit(1);
 184 +    }
 185 +
 186 +    poptFreeContext(optCon);
 187 +
 188 +    buffer = malloc( bufferSize );
 189 +
 190 +    int ret, getstatus;
 191 +    pico_Char * inp = NULL;
 192 +    pico_Char * local_text = NULL;
 193 +    short       outbuf[MAX_OUTBUF_SIZE/2];
 194 +    pico_Int16  bytes_sent, bytes_recv, text_remaining, out_data_type;
 195 +    pico_Retstring outMessage;
 196 +
 197 +    picoSynthAbort = 0;
 198 +
 199 +    picoMemArea = malloc( PICO_MEM_SIZE );
 200 +    if((ret = pico_initialize( picoMemArea, PICO_MEM_SIZE, &picoSystem ))) {
 201 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 202 +        fprintf(stderr, "Cannot initialize pico (%i): %s\n", ret, outMessage);
 203 +        goto terminate;
 204 +    }
 205 +
 206 +    /* Load the text analysis Lingware resource file.   */
 207 +    picoTaFileName      = (pico_Char *) malloc( PICO_MAX_DATAPATH_NAME_SIZE + PICO_MAX_FILE_NAME_SIZE );
 208 +    strcpy((char *) picoTaFileName,   PICO_LINGWARE_PATH);
 209 +    strcat((char *) picoTaFileName,   (const char *) picoInternalTaLingware[langIndex]);
 210 +    if((ret = pico_loadResource( picoSystem, picoTaFileName, &picoTaResource ))) {
 211 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 212 +        fprintf(stderr, "Cannot load text analysis resource file (%i): %s\n", ret, outMessage);
 213 +        goto unloadTaResource;
 214 +    }
 215 +
 216 +    /* Load the signal generation Lingware resource file.   */
 217 +    picoSgFileName      = (pico_Char *) malloc( PICO_MAX_DATAPATH_NAME_SIZE + PICO_MAX_FILE_NAME_SIZE );
 218 +    strcpy((char *) picoSgFileName,   PICO_LINGWARE_PATH);
 219 +    strcat((char *) picoSgFileName,   (const char *) picoInternalSgLingware[langIndex]);
 220 +    if((ret = pico_loadResource( picoSystem, picoSgFileName, &picoSgResource ))) {
 221 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 222 +        fprintf(stderr, "Cannot load signal generation Lingware resource file (%i): %s\n", ret, outMessage);
 223 +        goto unloadSgResource;
 224 +    }
 225 +
 226 +    /* Load the utpp Lingware resource file if exists - NOTE: this file is optional
 227 +       and is currently not used. Loading is only attempted for future compatibility.
 228 +       If this file is not present the loading will still succeed.                      //
 229 +    picoUtppFileName      = (pico_Char *) malloc( PICO_MAX_DATAPATH_NAME_SIZE + PICO_MAX_FILE_NAME_SIZE );
 230 +    strcpy((char *) picoUtppFileName,   PICO_LINGWARE_PATH);
 231 +    strcat((char *) picoUtppFileName,   (const char *) picoInternalUtppLingware[langIndex]);
 232 +    ret = pico_loadResource( picoSystem, picoUtppFileName, &picoUtppResource );
 233 +    pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 234 +    printf("pico_loadResource: %i: %s\n", ret, outMessage);
 235 +    */
 236 +
 237 +    /* Get the text analysis resource name.     */
 238 +    picoTaResourceName  = (pico_Char *) malloc( PICO_MAX_RESOURCE_NAME_SIZE );
 239 +    if((ret = pico_getResourceName( picoSystem, picoTaResource, (char *) picoTaResourceName ))) {
 240 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 241 +        fprintf(stderr, "Cannot get the text analysis resource name (%i): %s\n", ret, outMessage);
 242 +        goto unloadUtppResource;
 243 +    }
 244 +
 245 +    /* Get the signal generation resource name. */
 246 +    picoSgResourceName  = (pico_Char *) malloc( PICO_MAX_RESOURCE_NAME_SIZE );
 247 +    if((ret = pico_getResourceName( picoSystem, picoSgResource, (char *) picoSgResourceName ))) {
 248 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 249 +        fprintf(stderr, "Cannot get the signal generation resource name (%i): %s\n", ret, outMessage);
 250 +        goto unloadUtppResource;
 251 +    }
 252 +
 253 +
 254 +    /* Create a voice definition.   */
 255 +    if((ret = pico_createVoiceDefinition( picoSystem, (const pico_Char *) PICO_VOICE_NAME ))) {
 256 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 257 +        fprintf(stderr, "Cannot create voice definition (%i): %s\n", ret, outMessage);
 258 +        goto unloadUtppResource;
 259 +    }
 260 +
 261 +    /* Add the text analysis resource to the voice. */
 262 +    if((ret = pico_addResourceToVoiceDefinition( picoSystem, (const pico_Char *) PICO_VOICE_NAME, picoTaResourceName ))) {
 263 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 264 +        fprintf(stderr, "Cannot add the text analysis resource to the voice (%i): %s\n", ret, outMessage);
 265 +        goto unloadUtppResource;
 266 +    }
 267 +
 268 +    /* Add the signal generation resource to the voice. */
 269 +    if((ret = pico_addResourceToVoiceDefinition( picoSystem, (const pico_Char *) PICO_VOICE_NAME, picoSgResourceName ))) {
 270 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 271 +        fprintf(stderr, "Cannot add the signal generation resource to the voice (%i): %s\n", ret, outMessage);
 272 +        goto unloadUtppResource;
 273 +    }
 274 +
 275 +    /* Create a new Pico engine. */
 276 +    if((ret = pico_newEngine( picoSystem, (const pico_Char *) PICO_VOICE_NAME, &picoEngine ))) {
 277 +        pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 278 +        fprintf(stderr, "Cannot create a new pico engine (%i): %s\n", ret, outMessage);
 279 +        goto disposeEngine;
 280 +    }
 281 +
 282 +    local_text = (pico_Char *) text ;
 283 +    text_remaining = strlen((const char *) local_text) + 1;
 284 +
 285 +    inp = (pico_Char *) local_text;
 286 +
 287 +    size_t bufused = 0;
 288 +
 289 +    picoos_Common common = (picoos_Common) pico_sysGetCommon(picoSystem);
 290 +
 291 +    picoos_SDFile sdOutFile = NULL;
 292 +
 293 +    picoos_bool done = TRUE;
 294 +    if(TRUE != (done = picoos_sdfOpenOut(common, &sdOutFile,
 295 +        (picoos_char *) wavefile, SAMPLE_FREQ_16KHZ, PICOOS_ENC_LIN)))
 296 +    {
 297 +        fprintf(stderr, "Cannot open output wave file\n");
 298 +        ret = 1;
 299 +        goto disposeEngine;
 300 +    }
 301 +
 302 +    /* synthesis loop   */
 303 +    while (text_remaining) {
 304 +        /* Feed the text into the engine.   */
 305 +        if((ret = pico_putTextUtf8( picoEngine, inp, text_remaining, &bytes_sent ))) {
 306 +            pico_getSystemStatusMessage(picoSystem, ret, outMessage);
 307 +            fprintf(stderr, "Cannot put Text (%i): %s\n", ret, outMessage);
 308 +            goto disposeEngine;
 309 +        }
 310 +
 311 +        text_remaining -= bytes_sent;
 312 +        inp += bytes_sent;
 313 +
 314 +        do {
 315 +            if (picoSynthAbort) {
 316 +                goto disposeEngine;
 317 +            }
 318 +            /* Retrieve the samples and add them to the buffer. */
 319 +            getstatus = pico_getData( picoEngine, (void *) outbuf,
 320 +                      MAX_OUTBUF_SIZE, &bytes_recv, &out_data_type );
 321 +            if((getstatus !=PICO_STEP_BUSY) && (getstatus !=PICO_STEP_IDLE)){
 322 +                pico_getSystemStatusMessage(picoSystem, getstatus, outMessage);
 323 +                fprintf(stderr, "Cannot get Data (%i): %s\n", getstatus, outMessage);
 324 +                goto disposeEngine;
 325 +            }
 326 +            if (bytes_recv) {
 327 +                if ((bufused + bytes_recv) <= bufferSize) {
 328 +                    memcpy(buffer+bufused, (int8_t *) outbuf, bytes_recv);
 329 +                    bufused += bytes_recv;
 330 +                } else {
 331 +                    done = picoos_sdfPutSamples(
 332 +                                        sdOutFile,
 333 +                                        bufused / 2,
 334 +                                        (picoos_int16*) (buffer));
 335 +                    bufused = 0;
 336 +                    memcpy(buffer, (int8_t *) outbuf, bytes_recv);
 337 +                    bufused += bytes_recv;
 338 +                }
 339 +            }
 340 +        } while (PICO_STEP_BUSY == getstatus);
 341 +        /* This chunk of synthesis is finished; pass the remaining samples. */
 342 +        if (!picoSynthAbort) {
 343 +                    done = picoos_sdfPutSamples(
 344 +                                        sdOutFile,
 345 +                                        bufused / 2,
 346 +                                        (picoos_int16*) (buffer));
 347 +        }
 348 +        picoSynthAbort = 0;
 349 +    }
 350 +
 351 +    if(TRUE != (done = picoos_sdfCloseOut(common, &sdOutFile)))
 352 +    {
 353 +        fprintf(stderr, "Cannot close output wave file\n");
 354 +        ret = 1;
 355 +        goto disposeEngine;
 356 +    }
 357 +
 358 +disposeEngine:
 359 +    if (picoEngine) {
 360 +        pico_disposeEngine( picoSystem, &picoEngine );
 361 +        pico_releaseVoiceDefinition( picoSystem, (pico_Char *) PICO_VOICE_NAME );
 362 +        picoEngine = NULL;
 363 +    }
 364 +unloadUtppResource:
 365 +    if (picoUtppResource) {
 366 +        pico_unloadResource( picoSystem, &picoUtppResource );
 367 +        picoUtppResource = NULL;
 368 +    }
 369 +unloadSgResource:
 370 +    if (picoSgResource) {
 371 +        pico_unloadResource( picoSystem, &picoSgResource );
 372 +        picoSgResource = NULL;
 373 +    }
 374 +unloadTaResource:
 375 +    if (picoTaResource) {
 376 +        pico_unloadResource( picoSystem, &picoTaResource );
 377 +        picoTaResource = NULL;
 378 +    }
 379 +terminate:
 380 +    if (picoSystem) {
 381 +        pico_terminate(&picoSystem);
 382 +        picoSystem = NULL;
 383 +    }
 384 +    exit(ret);
 385 +}
 386 +
 387 diff --git a/pico/configure.in b/pico/configure.in
 388 index 0afb56d..349eb1d 100644
 389 --- a/pico/configure.in
 390 +++ b/pico/configure.in
 391 @@ -14,3 +14,6 @@ AC_CONFIG_FILES([Makefile])
 392  AC_OUTPUT
 393
 394  AC_CONFIG_MACRO_DIR([m4])
 395 +
 396 +AC_CHECK_LIB(popt, poptGetContext)
 397 +
 398 --
 399 1.7.1
 400