bool stardict_tts_plugin_init(StarDictTtsPlugInObject *obj)
{
flite_init();
register_cmu_us_kal(NULL);
register_cmu_time_awb(NULL);
register_cmu_us_kal16(NULL);
register_cmu_us_awb(NULL);
register_cmu_us_rms(NULL);
register_cmu_us_slt(NULL);
std::string res = get_cfg_filename();
if (!g_file_test(res.c_str(), G_FILE_TEST_EXISTS)) {
g_file_set_contents(res.c_str(), "[flite]\nvoice=\n", -1, NULL);
}
GKeyFile *keyfile = g_key_file_new();
g_key_file_load_from_file(keyfile, res.c_str(), G_KEY_FILE_NONE, NULL);
gchar *str = g_key_file_get_string(keyfile, "flite", "voice", NULL);
g_key_file_free(keyfile);
if (str) {
voice_engine = str;
g_free(str);
}
if (!voice_engine.empty()) {
now_voice = flite_voice_select(voice_engine.c_str());
} else {
now_voice = flite_voice_select(NULL);
}
obj->saytext_func = saytext;
obj->tts_name = _("Flite TTS");
g_print(_("Flite plug-in loaded.\n"));
return false;
}
int main(int argc, char** argv)
{
flite_init();
v=register_cmu_us_kal16();
flite_text_to_speech("Hello world",v,"play");
printf ("Finished \n");
exit(0);
}
int main(int argc, char **argv)
{
cst_voice *v;
const char *filename;
const char *outtype;
int i;
float durs;
int flite_verbose, flite_loop, flite_bench;
int explicit_filename, explicit_text, explicit_phones;
#define ITER_MAX 3
int bench_iter = 0;
cst_features *extra_feats;
InitThreads(HT_NOMONITOR); // enable msg driven monitoring
if(InitShell(argc,argv,version)<SUCCESS) return FAIL;
InitMem(); InitLabel();
InitMath(); InitSigP(); InitUtil();
InitWave(); InitAudio();
filename = 0;
outtype = "play"; /* default is to play */
flite_verbose = FALSE;
flite_loop = FALSE;
flite_bench = FALSE;
explicit_text = explicit_filename = explicit_phones = FALSE;
extra_feats = new_features();
flite_init();
for (i=1; i<argc; i++)
{
if (cst_streq(argv[i],"--version"))
{
flite_version();
return 1;
}
else if (cst_streq(argv[i],"-h") ||
cst_streq(argv[i],"--help") ||
cst_streq(argv[i],"-?"))
flite_usage();
else if (cst_streq(argv[i],"-v"))
flite_verbose = TRUE;
else if (cst_streq(argv[i],"-l"))
flite_loop = TRUE;
else if (cst_streq(argv[i],"-b"))
{
flite_bench = TRUE;
break; /* ignore other arguments */
}
else if ((cst_streq(argv[i],"-o")) && (i+1 < argc))
{
outtype = argv[i+1];
i++;
}
else if (cst_streq(argv[i],"-f") && (i+1 < argc))
{
filename = argv[i+1];
explicit_filename = TRUE;
i++;
}
else if ((cst_streq(argv[i],"-set") || cst_streq(argv[i],"-s"))
&& (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],0);
i++;
}
else if (cst_streq(argv[i],"--seti") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"int");
i++;
}
else if (cst_streq(argv[i],"--setf") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"float");
i++;
}
else if (cst_streq(argv[i],"--sets") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"string");
i++;
}
else if (cst_streq(argv[i],"-p") && (i+1 < argc))
{
filename = argv[i+1];
explicit_phones = TRUE;
i++;
}
else if (cst_streq(argv[i],"-t") && (i+1 < argc))
{
filename = argv[i+1];
explicit_text = TRUE;
i++;
}
else if (filename)
outtype = argv[i];
else
filename = argv[i];
}
if (filename == NULL) filename = "-"; /* stdin */
v = register_cmu_us_kal16(NULL);
feat_copy_into(extra_feats,v->features);
durs = 0.0;
if (flite_bench)
{
outtype = "none";
filename = "A whole joy was reaping, but they've gone south, you should fetch azure mike.";
explicit_text = TRUE;
}
loop:
if (explicit_phones)
durs = flite_phones_to_speech(filename,v,outtype);
else if ((strchr(filename,' ') && !explicit_filename) || explicit_text)
durs = flite_text_to_speech(filename,v,outtype);
else
durs = flite_file_to_speech(filename,v,outtype);
if (flite_loop || (flite_bench && bench_iter++ < ITER_MAX))
goto loop;
delete_features(extra_feats);
unregister_cmu_us_kal16(v);
return 0;
}
int main(int argc, char *argv[])
{
char *s,*fn;
cst_voice *voice; // synthesis voice
cst_utterance *utt; // current utterance
cst_wave *cstwave; // synthesised wave
Wave w; // HTK wave
short *p;
HTime sampPeriod = 625.0;
int n;
MemHeap mem;
AudioOut a;
try {
if (InitHTK(argc,argv,version)<SUCCESS){
ReportErrors("Main",0); exit(-1);
}
if (NumArgs() !=2) {
printf("SFliteTest synthstring file\n"); exit(0);
}
CreateHeap(&mem,"heap",MSTAK,1,0.0,10000,100000);
s = GetStrArg();
fn = GetStrArg();
printf("Synth: %s -> %s\n",s,fn);
// initialise Edinburgh cst lib
cst_regex_init();
// setup the voice
voice = register_cmu_us_kal16(NULL);
// convert text to waveform
utt = flite_synth_text(s,voice);
if (utt==NULL) {
HRError(12001,"SFliteTest: cant synthesise %s\n",s);
throw ATK_Error(12001);
}
cstwave = utt_wave(utt);
p = cstwave->samples; n = cstwave->num_samples;
w = OpenWaveOutput(&mem,&sampPeriod,n);
printf("%d samples created\n",n);
PutWaveSample(w,n,p);
if (CloseWaveOutput(w,WAV,fn)<SUCCESS){
ReportErrors("Main",0); exit(-1);
}
// explore structure
const cst_item *it, *itlast = NULL;
float x,y;
int i;
string lastword="0"; x = 0;
for (i=1,it = relation_head(utt_relation(utt, "Segment")); it!=NULL; it = item_next(it),i++)
{
printf("Segment %d\n",i);
y = item_feat_float(it,"end");
string ph = string(ffeature_string(it,"p.name"));
string wd = string(ffeature_string(it,"R:SylStructure.parent.parent.name"));
//printf("end = %f ph=%s wd=%s\n",y,ph.c_str(),wd.c_str());
if (wd != lastword){
printf("**** end of %s = %f\n",lastword.c_str(),x);
lastword=wd;
}
x = y;
}
//if (itlast!=NULL) {
// word = string(ffeature_string(itlast,"R:SylStructure.parent.parent.name"));
// idx = text.find(word);
//}
return 0;
}
catch (ATK_Error e){ ReportErrors("ATK",e.i); }
catch (HTK_Error e){ ReportErrors("HTK",e.i); }
return 0;
}
