void Reader::ReadNextLine( Book& book )
{
string strValue = book.GetMostRecentLine();
cout << endl << "SAY LINE " << endl << strValue << endl;
flite_text_to_speech( strValue.c_str(), m_ptrVoice, "play" );
m_currentLine++;
}
static PyObject* PySemsAudioFile_tts(PyObject* cls, PyObject* args)
{
char* text;
if(!PyArg_ParseTuple(args,"s",&text))
return NULL;
PyObject* constr_args = Py_BuildValue("(O)",Py_None);
PyObject* tts_file = PyObject_CallObject(cls,constr_args);
Py_DECREF(constr_args);
if(tts_file == NULL) {
PyErr_Print();
PyErr_SetString(PyExc_RuntimeError,"could not create new PySemsAudioFile object");
return NULL;
}
PySemsAudioFile* self = (PySemsAudioFile*)tts_file;
*self->filename = string(TTS_CACHE_PATH) + AmSession::getNewId() + string(".wav");
self->del_file = true;
flite_text_to_speech(text,self->tts_voice,self->filename->c_str());
if(self->af->open(self->filename->c_str(),AmAudioFile::Read)) {
Py_DECREF(tts_file);
PyErr_SetString(PyExc_IOError,"could not open TTS file");
return NULL;
}
return tts_file;
}
static void saytext(const char *text)
{
if (now_voice == NULL) {
now_voice = flite_voice_select(NULL);
}
if (now_voice) {
flite_text_to_speech(text, now_voice, "play");
}
}
int APIENTRY _tWinMain(HINSTANCE hInstance,
HINSTANCE hPrevInstance,
LPTSTR lpCmdLine,
int nCmdShow)
{
flite_init();
cst_voice *v1 = register_cmu_us_rms(NULL);
cst_voice *v2 = register_cmu_us_slt(NULL);
for(int i = 0; i < 3; ++i)
{
flite_text_to_speech("Welcome to the testing facility", v1, "play");
::Beep( 1000, 800);
flite_text_to_speech("Welcome to the testing facility", v2, "play");
::Beep( 1000, 800);
}
}
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)
{
flite_init();
v=register_cmu_time_awb();
flite_text_to_speech("11:52 PM",v,"flite_test.wav");
printf ("Finished \n");
exit(0);
}
void *tell_us_the_position(void *threadid)
{
cst_voice *v;
char choise[2];
char text_to_speech[MAX_TEXT_TO_SPEECH];
unsigned int current_position;
signed int degree_difference;
signed char pich,roll;
unsigned char should_speaks_the_route = 0;
// Init TTS system
flite_init();
// Choose language
v=register_irstifd_ita_pc();
sprintf(text_to_speech,"Fred, versione %s",VERSION);
flite_text_to_speech(text_to_speech,v,"play");
flite_text_to_speech("Il sistema è pronto",v,"play");
while(1)
{
int ret = read(PIPE_CHANNEL[0], choise, 1);
if(ret > 0)
{
read_compass_global_var(¤t_position,&pich,&roll);
if(choise[0] == '1') sprintf(text_to_speech,"Posizione %i",current_position);
if(choise[0] == '2') sprintf(text_to_speech,"Beccheggio %i",pich);
if(choise[0] == '3') sprintf(text_to_speech,"Rollio %i",roll);
if(choise[0] == '4')
{
sprintf(text_to_speech,"Posizione di partenza impostanta a %i",current_position);
START_POSITION = current_position;
should_speaks_the_route = 1;
}
// Exit request
if(choise[0] == '5') break;
// Text to speech
flite_text_to_speech(text_to_speech,v,"play");
}
// Check for periodic announcement
if(should_speaks_the_route == 1)
{
if(is_time_to_speak())
{
degree_difference = speaks_route();
if(degree_difference != 0)
{
sprintf(text_to_speech,"Fuori rotta di %i",degree_difference);
flite_text_to_speech(text_to_speech,v,"play");
}
}
}
}
flite_text_to_speech("Richiesta uscita programma",v,"play");
flite_text_to_speech("Fred vi saluta, ciao",v,"play");
// End thread
pthread_exit(NULL);
}
static void speak_dataset_comments (const dataset *dset)
{
int i;
cst_voice *v;
flite_init();
v = register_cmu_us_kal();
for (i=0; i<N_COMMENTS; i++) {
if (dset->comments[i] != NULL) {
flite_text_to_speech(dset->comments[i], v, "play");
}
}
}
static int speak_buffer (const char *buf, int (*should_stop)())
{
cst_voice *v;
char line[2048];
flite_init();
v = register_cmu_us_kal();
bufgets_init(buf);
while (bufgets(line, sizeof line, buf)) {
if (should_stop()) {
flite_text_to_speech("OK, stopping", v, "play");
break;
}
tailstrip(line);
flite_text_to_speech(line, v, "play");
}
bufgets_finalize(buf);
return 0;
}
static int speak_line (const char *line)
{
static cst_voice *v;
if (v == NULL) {
flite_init();
v = register_cmu_us_kal();
}
if (v == NULL) return 1;
flite_text_to_speech(line, v, "play");
return 0;
}
void ConferenceDialog::sayTTS(string text) {
string filename = string(TTS_CACHE_PATH) + text /* AmSession::getNewId() */
+ string(".wav");
last_sas = text;
flite_text_to_speech(text.c_str(),tts_voice,filename.c_str());
AmAudioFile* af = new AmAudioFile();
if(!af->open(filename.c_str(), AmAudioFile::Read)) {
play_list.addToPlayListFront(new AmPlaylistItem(af, NULL));
TTSFiles.push_back(af);
} else {
ERROR("ERROR reading TTSed file %s\n", filename.c_str());
delete af;
}
}
int main(int argc, char **argv)
{
cst_voice *v;
char thetime[1024];
char b[3];
int hour, min;
cst_regex *timex;
char *output = "play";
if (argc != 2)
{
fprintf(stderr,"usage: flite_time HH:MM\n");
exit(-1);
}
timex = new_cst_regex("[012][0-9]:[0-5][0-9]");
if (!cst_regex_match(timex,argv[1]))
{
fprintf(stderr,"not a valid time\n");
fprintf(stderr,"usage: flite_time HH:MM\n");
exit(-1);
}
delete_cst_regex(timex);
b[2] = '\0';
b[0] = argv[1][0];
b[1] = argv[1][1];
hour = atoi(b);
b[0] = argv[1][3];
b[1] = argv[1][4];
min = atoi(b);
flite_init();
v = register_cmu_time_awb(NULL);
sprintf(thetime,
"The time is now, %s %s %s, %s.",
time_approx(hour,min),
time_min(hour,min),
time_hour(hour,min),
time_tod(hour,min));
printf("%s\n",thetime);
flite_text_to_speech(thetime,v,output);
return 0;
}
static void audio_graph_error (const char *msg)
{
#ifdef HAVE_FLITE
cst_voice *v;
#endif
#ifdef WIN32_SAPI
ISpVoice *v = NULL;
HRESULT hr;
#endif
fprintf(stderr, "%s\n", msg);
#ifdef HAVE_FLITE
flite_init();
v = register_cmu_us_kal();
flite_text_to_speech(msg, v, "play");
#endif
#ifdef WIN32_SAPI
hr = CoInitialize(NULL);
if (!SUCCEEDED(hr)) return;
hr = CoCreateInstance(&CLSID_SpVoice,
NULL,
CLSCTX_ALL,
&IID_ISpVoice,
(void *) &v);
if (SUCCEEDED(hr)) {
wchar_t *w = wide_string(msg);
ISpVoice_Speak(v, w, 0, NULL);
free(w);
ISpVoice_Release(v);
}
#endif
}
int main(int argc, char **argv)
{
struct timeval tv;
cst_voice *v;
const char *filename;
const char *outtype;
int i;
float durs;
double time_start, time_end;
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;
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],"-pw"))
{
feat_set_string(extra_feats,"print_info_relation","Word");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ps"))
{
feat_set_string(extra_feats,"print_info_relation","Segment");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-pr") && (i+1 < argc))
{
feat_set_string(extra_feats,"print_info_relation",argv[i+1]);
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
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_VOX(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:
gettimeofday(&tv,NULL);
time_start = (double)(tv.tv_sec)+(((double)tv.tv_usec)/1000000.0);
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);
gettimeofday(&tv,NULL);
time_end = ((double)(tv.tv_sec))+((double)tv.tv_usec/1000000.0);
if (flite_verbose || (flite_bench && bench_iter == ITER_MAX))
printf("times faster than real-time: %f\n(%f seconds of speech synthesized in %f)\n",
durs/(float)(time_end-time_start),
durs,
(float)(time_end-time_start));
if (flite_loop || (flite_bench && bench_iter++ < ITER_MAX))
goto loop;
delete_features(extra_feats);
UNREGISTER_VOX(v);
return 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)
{
struct timeval tv;
cst_voice *v;
const char *filename;
const char *outtype;
cst_voice *desired_voice = 0;
const char *voicedir = NULL;
int i;
float durs;
double time_start, time_end;
int flite_verbose, flite_loop, flite_bench;
int explicit_filename, explicit_text, explicit_phones, ssml_mode;
#define ITER_MAX 3
int bench_iter = 0;
cst_features *extra_feats;
const char *lex_addenda_file = NULL;
const char *voicedumpfile = NULL;
cst_audio_streaming_info *asi;
filename = 0;
outtype = "play"; /* default is to play */
flite_verbose = FALSE;
flite_loop = FALSE;
flite_bench = FALSE;
explicit_text = explicit_filename = explicit_phones = FALSE;
ssml_mode = FALSE;
extra_feats = new_features();
flite_init();
flite_add_lang("eng",usenglish_init,cmu_lex_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],"-lv"))
{
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
flite_voice_list_print();
exit(0);
}
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],"-voice")) && (i+1 < argc))
{
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
desired_voice = flite_voice_select(argv[i+1]);
i++;
}
else if ((cst_streq(argv[i],"-voicedir")) && (i+1 < argc))
{
voicedir = argv[i+1];
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
i++;
}
else if ((cst_streq(argv[i],"-add_lex")) && (i+1 < argc))
{
lex_addenda_file = 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],"-pw"))
{
feat_set_string(extra_feats,"print_info_relation","Word");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ps"))
{
feat_set_string(extra_feats,"print_info_relation","Segment");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ssml"))
{
ssml_mode = TRUE;
}
else if (cst_streq(argv[i],"-pr") && (i+1 < argc))
{
feat_set_string(extra_feats,"print_info_relation",argv[i+1]);
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
i++;
}
else if (cst_streq(argv[i],"-voicedump") && (i+1 < argc))
{
voicedumpfile = argv[i+1];
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 */
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
if (desired_voice == 0)
desired_voice = flite_voice_select(NULL);
v = desired_voice;
feat_copy_into(extra_feats,v->features);
durs = 0.0;
if (voicedumpfile != NULL)
{
flite_voice_dump(v,voicedumpfile);
exit(0);
}
if (lex_addenda_file)
flite_voice_add_lex_addenda(v,lex_addenda_file);
if (cst_streq("stream",outtype))
{
asi = new_audio_streaming_info();
asi->asc = audio_stream_chunk;
feat_set(v->features,"streaming_info",audio_streaming_info_val(asi));
}
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:
gettimeofday(&tv,NULL);
time_start = (double)(tv.tv_sec)+(((double)tv.tv_usec)/1000000.0);
if (explicit_phones)
durs = flite_phones_to_speech(filename,v,outtype);
else if ((strchr(filename,' ') && !explicit_filename) || explicit_text)
{
if (ssml_mode)
durs = flite_ssml_text_to_speech(filename,v,outtype);
else
durs = flite_text_to_speech(filename,v,outtype);
}
else
{
if (ssml_mode)
durs = flite_ssml_file_to_speech(filename,v,outtype);
else
durs = flite_file_to_speech(filename,v,outtype);
}
gettimeofday(&tv,NULL);
time_end = ((double)(tv.tv_sec))+((double)tv.tv_usec/1000000.0);
if (flite_verbose || (flite_bench && bench_iter == ITER_MAX))
printf("times faster than real-time: %f\n(%f seconds of speech synthesized in %f)\n",
durs/(float)(time_end-time_start),
durs,
(float)(time_end-time_start));
if (flite_loop || (flite_bench && bench_iter++ < ITER_MAX))
goto loop;
delete_features(extra_feats);
delete_val(flite_voice_list); flite_voice_list=0;
/* cst_alloc_debug_summary(); */
return 0;
}
