void TextToSpeech::load(
char *dn_mecab, char *fn_ms_dur, char *fn_ts_dur,
char *fn_ms_mgc, char *fn_ts_mgc, char **fn_ws_mgc, int num_ws_mgc,
char *fn_ms_lf0, char *fn_ts_lf0, char **fn_ws_lf0, int num_ws_lf0,
char *fn_ms_lpf, char *fn_ts_lpf, char **fn_ws_lpf, int num_ws_lpf,
char *fn_ms_gvm, char *fn_ts_gvm, char *fn_ms_gvl, char *fn_ts_gvl,
char *fn_ms_gvf, char *fn_ts_gvf, char *fn_gv_switch)
{
Mecab_load(&open_jtalk_.mecab, dn_mecab);
HTS_Engine_load_duration_from_fn(&open_jtalk_.engine, &fn_ms_dur, &fn_ts_dur, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_mgc, &fn_ts_mgc, fn_ws_mgc, 0, FALSE, num_ws_mgc, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_lf0, &fn_ts_lf0, fn_ws_lf0, 1, TRUE, num_ws_lf0, 1);
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3)
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine, &fn_ms_lpf, &fn_ts_lpf, fn_ws_lpf, 2, FALSE, num_ws_lpf, 1);
if (fn_ms_gvm != NULL) {
if (fn_ts_gvm != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvm, &fn_ts_gvm, 0, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvm, NULL, 0, 1);
}
if (fn_ms_gvl != NULL) {
if (fn_ts_gvl != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvl, &fn_ts_gvl, 1, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvl, NULL, 1, 1);
}
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3 && fn_ms_gvf != NULL) {
if (fn_ts_gvf != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvf, &fn_ts_gvf, 2, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine, &fn_ms_gvf, NULL, 2, 1);
}
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&open_jtalk_.engine, fn_gv_switch);
}
void TextToSpeech::load(
const char *dn_mecab, const char *fn_ms_dur, const char *fn_ts_dur,
const char *fn_ms_mgc, const char *fn_ts_mgc, const char **fn_ws_mgc, int num_ws_mgc,
const char *fn_ms_lf0, const char *fn_ts_lf0, const char **fn_ws_lf0, int num_ws_lf0,
const char *fn_ms_lpf, const char *fn_ts_lpf, const char **fn_ws_lpf, int num_ws_lpf,
const char *fn_ms_gvm, const char *fn_ts_gvm, const char *fn_ms_gvl, const char *fn_ts_gvl,
const char *fn_ms_gvf, const char *fn_ts_gvf, const char *fn_gv_switch)
{
Mecab_load(&open_jtalk_.mecab, const_cast<char *>(dn_mecab));
HTS_Engine_load_duration_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_dur),
const_cast<char **>(&fn_ts_dur), 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_mgc),
const_cast<char **>(&fn_ts_mgc),
const_cast<char **>(fn_ws_mgc), 0, FALSE, num_ws_mgc, 1);
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_lf0),
const_cast<char **>(&fn_ts_lf0),
const_cast<char **>(fn_ws_lf0), 1, TRUE, num_ws_lf0, 1);
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3)
HTS_Engine_load_parameter_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_lpf),
const_cast<char **>(&fn_ts_lpf),
const_cast<char **>(fn_ws_lpf), 2, FALSE, num_ws_lpf, 1);
if (fn_ms_gvm != NULL) {
if (fn_ts_gvm != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvm),
const_cast<char **>(&fn_ts_gvm), 0, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvm), NULL, 0, 1);
}
if (fn_ms_gvl != NULL) {
if (fn_ts_gvl != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvl),
const_cast<char **>(&fn_ts_gvl), 1, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvl), NULL, 1, 1);
}
if (HTS_Engine_get_nstream(&open_jtalk_.engine) == 3 && fn_ms_gvf != NULL) {
if (fn_ts_gvf != NULL)
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvf),
const_cast<char **>(&fn_ts_gvf), 2, 1);
else
HTS_Engine_load_gv_from_fn(&open_jtalk_.engine,
const_cast<char **>(&fn_ms_gvf), NULL, 2, 1);
}
if (fn_gv_switch != NULL)
HTS_Engine_load_gv_switch_from_fn(&open_jtalk_.engine,
const_cast<char *>(fn_gv_switch));
}
static int setup(struct app *app)
{
#ifdef HTS_MELP
#define NR_STREAMS 3
#else
#define NR_STREAMS 2
#endif /* HTS_MELP */
double gv_weight[] = {
app->gv_weight_mgc,
app->gv_weight_lf0,
#ifdef HTS_MELP
app->gv_weight_lpf
#endif /* HTS_MELP */
};
int i;
app->play_h = play_init(&app->play_info, "default",
SND_PCM_FORMAT_S16_LE, 1, app->sampling_rate,
500000, 8);
Mecab_initialize(&app->mecab);
if (Mecab_load(&app->mecab, app->dn_mecab) != TRUE)
return -1;
NJD_initialize(&app->njd);
JPCommon_initialize(&app->jpcommon);
HTS_Engine_initialize(&app->engine);
if (HTS_Engine_load(&app->engine, &app->fn_voice, 1) != TRUE)
return -1;
HTS_Engine_set_sampling_frequency(&app->engine,
(size_t)app->sampling_rate);
if (app->fperiod >= 0)
HTS_Engine_set_fperiod(&app->engine, app->fperiod);
if (app->alpha >= 0.0)
HTS_Engine_set_alpha(&app->engine, app->alpha);
if (app->beta >= 0.0)
HTS_Engine_set_beta(&app->engine, app->beta);
if (app->half_tone >= 0.0)
HTS_Engine_add_half_tone(&app->engine, app->half_tone);
if (app->audio_buff_size > 0)
HTS_Engine_set_audio_buff_size(&app->engine,
app->audio_buff_size);
if (app->uv_threshold >= 0.0)
HTS_Engine_set_msd_threshold(&app->engine, 1,
app->uv_threshold);
if (app->speed >= 0.0)
HTS_Engine_set_speed(&app->engine, app->speed);
for (i = 0; i < NR_STREAMS; i++)
if (gv_weight[i] >= 0.0)
HTS_Engine_set_gv_weight(&app->engine, i, gv_weight[i]);
return 0;
}
static int Open_JTalk_load(Open_JTalk * open_jtalk, char *dn_mecab, char *fn_voice)
{
if (Mecab_load(&open_jtalk->mecab, dn_mecab) != TRUE) {
Open_JTalk_clear(open_jtalk);
return 0;
}
if (HTS_Engine_load(&open_jtalk->engine, &fn_voice, 1) != TRUE) {
Open_JTalk_clear(open_jtalk);
return 0;
}
return 1;
}
