void TextToSpeech::synthesis(char *txt, FILE * wavfp)
{
char buff[MAXBUFLEN];
text2mecab(buff, txt);
Mecab_analysis(&open_jtalk_.mecab, buff);
mecab2njd(&open_jtalk_.njd, Mecab_get_feature(&open_jtalk_.mecab), Mecab_get_size(&open_jtalk_.mecab));
njd_set_pronunciation(&open_jtalk_.njd);
njd_set_digit(&open_jtalk_.njd);
njd_set_accent_phrase(&open_jtalk_.njd);
njd_set_accent_type(&open_jtalk_.njd);
njd_set_unvoiced_vowel(&open_jtalk_.njd);
njd_set_long_vowel(&open_jtalk_.njd);
njd2jpcommon(&open_jtalk_.jpcommon, &open_jtalk_.njd);
JPCommon_make_label(&open_jtalk_.jpcommon);
if (JPCommon_get_label_size(&open_jtalk_.jpcommon) > 2) {
HTS_Engine_load_label_from_string_list(
&open_jtalk_.engine,
JPCommon_get_label_feature(&open_jtalk_.jpcommon),
JPCommon_get_label_size(&open_jtalk_.jpcommon)
);
HTS_Engine_create_sstream(&open_jtalk_.engine);
HTS_Engine_create_pstream(&open_jtalk_.engine);
HTS_Engine_create_gstream(&open_jtalk_.engine);
if (wavfp != NULL)
HTS_Engine_save_riff(&open_jtalk_.engine, wavfp);
HTS_Engine_refresh(&open_jtalk_.engine);
}
JPCommon_refresh(&open_jtalk_.jpcommon);
NJD_refresh(&open_jtalk_.njd);
Mecab_refresh(&open_jtalk_.mecab);
}
void TextToSpeech::synthesis(const char *txt)
{
char buff[MAXBUFLEN];
text2mecab(buff, txt);
Mecab_analysis(&open_jtalk_.mecab, buff);
mecab2njd(&open_jtalk_.njd, Mecab_get_feature(&open_jtalk_.mecab), Mecab_get_size(&open_jtalk_.mecab));
njd_set_pronunciation(&open_jtalk_.njd);
njd_set_digit(&open_jtalk_.njd);
njd_set_accent_phrase(&open_jtalk_.njd);
njd_set_accent_type(&open_jtalk_.njd);
njd_set_unvoiced_vowel(&open_jtalk_.njd);
njd_set_long_vowel(&open_jtalk_.njd);
njd2jpcommon(&open_jtalk_.jpcommon, &open_jtalk_.njd);
JPCommon_make_label(&open_jtalk_.jpcommon);
if (JPCommon_get_label_size(&open_jtalk_.jpcommon) > 2) {
unsigned int pcm_len;
HTS_Engine_load_label_from_string_list(
&open_jtalk_.engine,
JPCommon_get_label_feature(&open_jtalk_.jpcommon),
JPCommon_get_label_size(&open_jtalk_.jpcommon)
);
HTS_Engine_create_sstream(&open_jtalk_.engine);
HTS_Engine_create_pstream(&open_jtalk_.engine);
HTS_Engine_create_gstream(&open_jtalk_.engine);
pcm_len = HTS_Engine_get_generated_speech_size(&open_jtalk_.engine);
pcm_ = new short[pcm_len];
HTS_Engine_get_generated_speech(&open_jtalk_.engine, pcm_);
play_write(play_h_, pcm_, pcm_len * sizeof(short));
HTS_Engine_refresh(&open_jtalk_.engine);
}
JPCommon_refresh(&open_jtalk_.jpcommon);
NJD_refresh(&open_jtalk_.njd);
Mecab_refresh(&open_jtalk_.mecab);
}
int OpenJTalk_synthesis_towav(OpenJTalk** openjtalk,const char* text, const char* wavfilename)
{
char buff[MAXBUFLEN];
FILE * wavfp;
wavfp = fopen(wavfilename,"wb");
if (!wavfp)
{
sprintf((*openjtalk)->errorout,"can not open %s.",wavfilename);
return 0;
}
text2mecab(buff, (char*)text);
Mecab_analysis((*openjtalk)->mecab, buff);
mecab2njd(&(*openjtalk)->njd, Mecab_get_feature((*openjtalk)->mecab),
Mecab_get_size((*openjtalk)->mecab));
njd_set_pronunciation(&(*openjtalk)->njd);
njd_set_digit(&(*openjtalk)->njd);
njd_set_accent_phrase(&(*openjtalk)->njd);
njd_set_accent_type(&(*openjtalk)->njd);
njd_set_unvoiced_vowel(&(*openjtalk)->njd);
njd_set_long_vowel(&(*openjtalk)->njd);
njd2jpcommon(&(*openjtalk)->jpcommon, &(*openjtalk)->njd);
JPCommon_make_label(&(*openjtalk)->jpcommon);
if (JPCommon_get_label_size(&(*openjtalk)->jpcommon) > 2) {
HTS_Engine_load_label_from_string_list(&(*openjtalk)->engine,
JPCommon_get_label_feature(&(*openjtalk)->jpcommon),
JPCommon_get_label_size(&(*openjtalk)->jpcommon));
HTS_Engine_create_sstream(&(*openjtalk)->engine);
HTS_Engine_create_pstream(&(*openjtalk)->engine);
HTS_Engine_create_gstream(&(*openjtalk)->engine);
HTS_Engine_save_riff(&(*openjtalk)->engine, wavfp);
/*
if (wavfp != NULL)
HTS_Engine_save_riff(&(*openjtalk)->engine, wavfp);
if (logfp != NULL) {
fprintf(logfp, "[Text analysis result]\n");
NJD_fprint(&(*openjtalk)->njd, logfp);
fprintf(logfp, "\n[Output label]\n");
HTS_Engine_save_label(&(*openjtalk)->engine, logfp);
fprintf(logfp, "\n");
HTS_Engine_save_information(&(*openjtalk)->engine, logfp);
}
*/
HTS_Engine_refresh(&(*openjtalk)->engine);
}
JPCommon_refresh(&(*openjtalk)->jpcommon);
NJD_refresh(&(*openjtalk)->njd);
Mecab_refresh((*openjtalk)->mecab);
fclose(wavfp);
return 1;
}
static int synthesize(struct app *app, char *txt)
{
char buff[MAXBUFLEN];
int label_size;
int r = -1;
text2mecab(buff, txt);
Mecab_analysis(&app->mecab, buff);
mecab2njd(&app->njd, Mecab_get_feature(&app->mecab),
Mecab_get_size(&app->mecab));
njd_set_pronunciation(&app->njd);
njd_set_digit(&app->njd);
njd_set_accent_phrase(&app->njd);
njd_set_accent_type(&app->njd);
njd_set_unvoiced_vowel(&app->njd);
njd_set_long_vowel(&app->njd);
njd2jpcommon(&app->jpcommon, &app->njd);
JPCommon_make_label(&app->jpcommon);
label_size = JPCommon_get_label_size(&app->jpcommon);
if (label_size > 2) {
if (HTS_Engine_synthesize_from_strings(
&app->engine,
JPCommon_get_label_feature(&app->jpcommon),
label_size) == TRUE) {
unsigned int pcm_len;
r = 0; /* success */
pcm_len = HTS_Engine_get_generated_speech_size(
&app->engine);
app->pcm = malloc(pcm_len * sizeof(short));
HTS_Engine_get_generated_speech(&app->engine, app->pcm);
play_write(app->play_h, app->pcm,
pcm_len * sizeof(short));
}
if (app->logfp) {
fprintf(app->logfp, "[Text analysis result]\n");
NJD_fprint(&app->njd, app->logfp);
fprintf(app->logfp, "\n[Output label]\n");
HTS_Engine_save_label(&app->engine, app->logfp);
fprintf(app->logfp, "\n");
HTS_Engine_save_information(&app->engine, app->logfp);
}
HTS_Engine_refresh(&app->engine);
}
JPCommon_refresh(&app->jpcommon);
NJD_refresh(&app->njd);
Mecab_refresh(&app->mecab);
return r;
}
static int Open_JTalk_synthesis(Open_JTalk * open_jtalk, const char *txt, FILE * wavfp,
FILE * logfp)
{
int result = 0;
char buff[MAXBUFLEN];
text2mecab(buff, txt);
Mecab_analysis(&open_jtalk->mecab, buff);
mecab2njd(&open_jtalk->njd, Mecab_get_feature(&open_jtalk->mecab),
Mecab_get_size(&open_jtalk->mecab));
njd_set_pronunciation(&open_jtalk->njd);
njd_set_digit(&open_jtalk->njd);
njd_set_accent_phrase(&open_jtalk->njd);
njd_set_accent_type(&open_jtalk->njd);
njd_set_unvoiced_vowel(&open_jtalk->njd);
njd_set_long_vowel(&open_jtalk->njd);
njd2jpcommon(&open_jtalk->jpcommon, &open_jtalk->njd);
JPCommon_make_label(&open_jtalk->jpcommon);
if (JPCommon_get_label_size(&open_jtalk->jpcommon) > 2) {
if (HTS_Engine_synthesize_from_strings
(&open_jtalk->engine, JPCommon_get_label_feature(&open_jtalk->jpcommon),
JPCommon_get_label_size(&open_jtalk->jpcommon)) == TRUE)
result = 1;
if (wavfp != NULL)
HTS_Engine_save_riff(&open_jtalk->engine, wavfp);
if (logfp != NULL) {
fprintf(logfp, "[Text analysis result]\n");
NJD_fprint(&open_jtalk->njd, logfp);
fprintf(logfp, "\n[Output label]\n");
HTS_Engine_save_label(&open_jtalk->engine, logfp);
fprintf(logfp, "\n");
HTS_Engine_save_information(&open_jtalk->engine, logfp);
}
HTS_Engine_refresh(&open_jtalk->engine);
}
JPCommon_refresh(&open_jtalk->jpcommon);
NJD_refresh(&open_jtalk->njd);
Mecab_refresh(&open_jtalk->mecab);
return result;
}
