/* Flite_HTS_Engine_load: load HTS voice */
HTS_Boolean Flite_HTS_Engine_load(Flite_HTS_Engine * f, const char *fn)
{
HTS_Boolean result;
char *voices = strdup(fn);
result = HTS_Engine_load(&f->engine, &voices, 1);
free(voices);
return result;
}
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;
}
void std_hts_engine_impl::do_initialize()
{
engine.reset(new HTS_Engine);
HTS_Engine_initialize(engine.get());
engine->audio.audio_interface=this;
std::string voice_path(path::join(data_path,"voice.data"));
char* c_voice_path=const_cast<char*>(voice_path.c_str());
if(!HTS_Engine_load(engine.get(),&c_voice_path,1))
{
HTS_Engine_clear(engine.get());
throw initialization_error();
}
HTS_Engine_set_beta(engine.get(),beta);
HTS_Engine_set_audio_buff_size(engine.get(),HTS_Engine_get_fperiod(engine.get()));
}
int main(int argc, char* argv[]) {
size_t num_voices;
char **fn_voices;
char* in_fname;
char* output_fname;
FILE * outfp;
char* dur_fname;
FILE * durfp;
bool print_label = false;
bool print_utt = false;
bool write_raw = false;
bool write_durlabel = false;
CFSAString LexFileName, LexDFileName;
HTS_Engine engine;
double speed = 1.1;
size_t fr = 48000;
size_t fp = 240;
float alpha = 0.55;
float beta = 0.0;
float ht = 2.0;
float th = 0.5;
float gvw1 = 1.0;
float gvw2 = 1.2;
FSCInit();
fn_voices = (char **) malloc(argc * sizeof (char *));
if (argc < 11) {
fprintf(stderr, "Viga: liiga vähe parameetreid\n\n");
PrintUsage();
}
for (int i = 0; i < argc; i++) {
if (CFSAString("-lex") == argv[i]) {
if (i + 1 < argc) {
LexFileName = argv[++i];
} else {
return PrintUsage();
}
}
if (CFSAString("-lexd") == argv[i]) {
if (i + 1 < argc) {
LexDFileName = argv[++i];
} else {
return PrintUsage();
}
}
if (CFSAString("-m") == argv[i]) {
if (i + 1 < argc) {
fn_voices[0] = argv[i + 1];
} else {
fprintf(stderr, "Viga: puudub *.htsvoice fail\n");
PrintUsage();
exit(0);
}
}
if (CFSAString("-o") == argv[i]) {
if (i + 1 < argc) {
output_fname = argv[i + 1];
cfileexists(output_fname);
} else {
fprintf(stderr, "Viga: puudb väljundfaili nimi\n");
PrintUsage();
exit(0);
}
}
if (CFSAString("-f") == argv[i]) {
if (i + 1 < argc) {
in_fname = argv[i + 1];
} else {
fprintf(stderr, "Viga: puudb sisendfaili nimi\n");
PrintUsage();
exit(0);
}
}
if (CFSAString("-s") == argv[i]) {
if (i + 1 < argc) {
samplerate(fr, fp, alpha, atoi(argv[i + 1]));
}
}
if (CFSAString("-r") == argv[i]) {
if (i + 1 < argc) {
speed = atof(argv[i + 1]);
}
}
if (CFSAString("-ht") == argv[i]) {
if (i + 1 < argc) {
ht = atof(argv[i + 1]);
}
}
if (CFSAString("-gvw1") == argv[i]) {
if (i + 1 < argc) {
gvw1 = atof(argv[i + 1]);
}
}
if (CFSAString("-gvw2") == argv[i]) {
if (i + 1 < argc) {
gvw2 = atof(argv[i + 1]);
}
}
if (CFSAString("-debug") == argv[i]) {
print_label = true;
}
if (CFSAString("-utt") == argv[i]) {
print_utt = true;
}
if (CFSAString("-raw") == argv[i]) {
write_raw = true;
}
if (CFSAString("-dur") == argv[i]) {
if (i + 1 < argc) {
dur_fname = argv[i + 1];
cfileexists(dur_fname);
write_durlabel = true;
} else {
fprintf(stderr, "Viga: puudb kestustefaili nimi\n");
PrintUsage();
exit(0);
}
}
}
Linguistic.Open(LexFileName);
Disambiguator.Open(LexDFileName);
CFSWString text;
ReadUTF8Text(text, in_fname);
HTS_Engine_initialize(&engine);
if (HTS_Engine_load(&engine, fn_voices, 1) != TRUE) {
fprintf(stderr, "Viga: puudub *.htsvoice. %p\n", fn_voices[0]);
free(fn_voices);
HTS_Engine_clear(&engine);
exit(1);
}
free(fn_voices);
HTS_Engine_set_sampling_frequency(&engine, (size_t) fr);
HTS_Engine_set_phoneme_alignment_flag(&engine, FALSE);
HTS_Engine_set_fperiod(&engine, (size_t) fp);
HTS_Engine_set_alpha(&engine, alpha);
HTS_Engine_set_beta(&engine, beta);
HTS_Engine_set_speed(&engine, speed);
HTS_Engine_add_half_tone(&engine, ht);
HTS_Engine_set_msd_threshold(&engine, 1, th);
/*
HTS_Engine_set_duration_interpolation_weight(&engine, 1, diw);
HTS_Engine_set_parameter_interpolation_weight(&engine, 0, 0, piw1);
HTS_Engine_set_parameter_interpolation_weight(&engine, 0, 1, piw2);
HTS_Engine_set_gv_interpolation_weight(&engine, 0, 0, giw1);
HTS_Engine_set_gv_interpolation_weight(&engine, 0, 1, giw2);
*/
HTS_Engine_set_gv_weight(&engine, 0, gvw1);
HTS_Engine_set_gv_weight(&engine, 1, gvw2);
text = DealWithText(text);
CFSArray<CFSWString> res = do_utterances(text);
INTPTR data_size = 0;
outfp = fopen(output_fname, "wb");
if (write_durlabel) durfp = fopen(dur_fname, "w");
if (!write_raw) HTS_Engine_write_header(&engine, outfp, 1);
for (INTPTR i = 0; i < res.GetSize(); i++) {
CFSArray<CFSWString> label = do_all(res[i], print_label, print_utt);
std::vector<std::string> v;
v = to_vector(label);
std::vector<char*> vc;
fill_char_vector(v, vc);
size_t n_lines = vc.size();
if (HTS_Engine_synthesize_from_strings(&engine, &vc[0], n_lines) != TRUE) {
fprintf(stderr, "Viga: süntees ebaonnestus.\n");
HTS_Engine_clear(&engine);
exit(1);
}
clean_char_vector(vc);
data_size += HTS_Engine_engine_speech_size(&engine);
if (write_durlabel) HTS_Engine_save_durlabel(&engine, durfp);
HTS_Engine_save_generated_speech(&engine, outfp);
HTS_Engine_refresh(&engine);
} //synth loop
if (!write_raw) HTS_Engine_write_header(&engine, outfp, data_size);
if (write_durlabel) fclose(durfp);
fclose(outfp);
HTS_Engine_clear(&engine);
Linguistic.Close();
FSCTerminate();
return 0;
}
int htsSynthesize(int argc, char **argv)
{
int i;
double f;
/* hts_engine API */
HTS_Engine engine;
/* HTS voices */
size_t num_voices;
char **fn_voices;
/* input label file name */
char *labfn = NULL;
/* output file pointers */
FILE *durfp = NULL, *mgcfp = NULL, *lf0fp = NULL, *lpffp = NULL, *wavfp = NULL, *rawfp = NULL, *tracefp = NULL;
/* interpolation weights */
size_t num_interpolation_weights;
/* output usage */
if (argc <= 1)
usage();
/* initialize hts_engine API */
HTS_Engine_initialize(&engine);
/* get HTS voice file names */
num_voices = 0;
fn_voices = (char **) malloc(argc * sizeof(char *));
for (i = 0; i < argc; i++)
{
if (argv[i][0] == '-' && argv[i][1] == 'm')
fn_voices[num_voices++] = argv[++i];
if (argv[i][0] == '-' && argv[i][1] == 'h')
usage();
}
if (num_voices == 0)
{
fprintf(stderr, "Error: HTS voice must be specified.\n");
free(fn_voices);
return (-1);
}
/* load HTS voices */
if (HTS_Engine_load(&engine, fn_voices, num_voices) != TRUE)
{
fprintf(stderr, "Error: HTS voices cannot be loaded.\n");
free(fn_voices);
HTS_Engine_clear(&engine);
return (-1);
}
free(fn_voices);
/* get options */
while (--argc)
{
if (**++argv == '-')
{
switch (*(*argv + 1))
{
case 'v':
switch (*(*argv + 2))
{
case 'p':
HTS_Engine_set_phoneme_alignment_flag(&engine, TRUE);
break;
default:
fprintf(stderr, "Error: Invalid option '-v%c'.\n", *(*argv + 2));
HTS_Engine_clear(&engine);
return (-1);
}
break;
case 'o':
switch (*(*argv + 2))
{
case 'w':
wavfp = fopen(*++argv, "wb");
break;
case 'r':
rawfp = fopen(*++argv, "wb");
break;
case 'd':
durfp = fopen(*++argv, "wt");
break;
case 'm':
mgcfp = fopen(*++argv, "wb");
break;
case 'f':
case 'p':
lf0fp = fopen(*++argv, "wb");
break;
case 'l':
lpffp = fopen(*++argv, "wb");
break;
case 't':
tracefp = fopen(*++argv, "wt");
break;
default:
fprintf(stderr, "Error: Invalid option '-o%c'.\n", *(*argv + 2));
HTS_Engine_clear(&engine);
return (-1);
}
--argc;
break;
case 'h':
usage();
break;
case 'm':
argv++; /* HTS voices were already loaded */
--argc;
break;
case 's':
HTS_Engine_set_sampling_frequency(&engine, (size_t) atoi(*++argv));
--argc;
break;
case 'p':
HTS_Engine_set_fperiod(&engine, (size_t) atoi(*++argv));
--argc;
break;
case 'a':
HTS_Engine_set_alpha(&engine, atof(*++argv));
--argc;
break;
case 'b':
HTS_Engine_set_beta(&engine, atof(*++argv));
--argc;
break;
case 'r':
HTS_Engine_set_speed(&engine, atof(*++argv));
--argc;
break;
case 'f':
switch (*(*argv + 2))
{
case 'm':
HTS_Engine_add_half_tone(&engine, atof(*++argv));
break;
default:
fprintf(stderr, "Error: Invalid option '-f%c'.\n", *(*argv + 2));
HTS_Engine_clear(&engine);
return (-1);
}
--argc;
break;
case 'u':
HTS_Engine_set_msd_threshold(&engine, 1, atof(*++argv));
--argc;
break;
case 'i':
num_interpolation_weights = atoi(*++argv);
argc--;
if (num_interpolation_weights != num_voices)
{
HTS_Engine_clear(&engine);
return(-1);
}
for (i = 0; i < (int) num_interpolation_weights; i++)
{
f = atof(*++argv);
argc--;
HTS_Engine_set_duration_interpolation_weight(&engine, i, f);
HTS_Engine_set_parameter_interpolation_weight(&engine, i, 0, f);
HTS_Engine_set_parameter_interpolation_weight(&engine, i, 1, f);
HTS_Engine_set_gv_interpolation_weight(&engine, i, 0, f);
HTS_Engine_set_gv_interpolation_weight(&engine, i, 1, f);
}
break;
case 'j':
switch (*(*argv + 2))
{
case 'm':
HTS_Engine_set_gv_weight(&engine, 0, atof(*++argv));
break;
case 'f':
case 'p':
HTS_Engine_set_gv_weight(&engine, 1, atof(*++argv));
break;
default:
fprintf(stderr, "Error: Invalid option '-j%c'.\n", *(*argv + 2));
HTS_Engine_clear(&engine);
return(-1);
}
--argc;
break;
case 'g':
HTS_Engine_set_volume(&engine, atof(*++argv));
--argc;
break;
case 'z':
HTS_Engine_set_audio_buff_size(&engine, (size_t) atoi(*++argv));
--argc;
break;
default:
fprintf(stderr, "Error: Invalid option '-%c'.\n", *(*argv + 1));
HTS_Engine_clear(&engine);
return(-1);
}
}
else
{
labfn = *argv;
}
}
/* synthesize */
if (HTS_Engine_synthesize_from_fn(&engine, labfn) != TRUE)
{
fprintf(stderr, "Error: waveform cannot be synthesized.\n");
HTS_Engine_clear(&engine);
return(-1);
}
/* output */
if (tracefp != NULL)
HTS_Engine_save_information(&engine, tracefp);
if (durfp != NULL)
HTS_Engine_save_label(&engine, durfp);
if (rawfp)
HTS_Engine_save_generated_speech(&engine, rawfp);
if (wavfp)
HTS_Engine_save_riff(&engine, wavfp);
if (mgcfp)
HTS_Engine_save_generated_parameter(&engine, 0, mgcfp);
if (lf0fp)
HTS_Engine_save_generated_parameter(&engine, 1, lf0fp);
if (lpffp)
HTS_Engine_save_generated_parameter(&engine, 2, lpffp);
/* reset */
HTS_Engine_refresh(&engine);
/* free memory */
HTS_Engine_clear(&engine);
/* close files */
if (durfp != NULL)
fclose(durfp);
if (mgcfp != NULL)
fclose(mgcfp);
if (lf0fp != NULL)
fclose(lf0fp);
if (lpffp != NULL)
fclose(lpffp);
if (wavfp != NULL)
fclose(wavfp);
if (rawfp != NULL)
fclose(rawfp);
if (tracefp != NULL)
fclose(tracefp);
return 0;
}
