static int save_dataset_comments (const dataset *dset)
{
int i, j;
cst_voice *v;
cst_wave *w, *fullw = NULL;
flite_init();
v = register_cmu_us_kal();
j = 0;
for (i=0; i<N_COMMENTS; i++) {
if (dset->comments[i] != NULL) {
if (j == 0) {
fullw = flite_text_to_wave(dset->comments[i], v);
} else {
w = flite_text_to_wave(dset->comments[i], v);
concat_wave(fullw, w);
delete_wave(w);
}
j++;
}
}
cst_wave_save_riff(fullw, "gretl_flite.wav");
delete_wave(fullw);
return 0;
}
int flowm_say_text(TCHAR *text)
{
char *s;
int ns;
cst_voice *v;
if (previous_wave)
{
delete_wave(previous_wave);
previous_wave = NULL;
}
s = cst_wstr2cstr(text); /* text to synthesize */
v = VoxDefs[flowm_selected_voice].v; /* voice to synthesize with */
feat_remove(v->features,"print_info_relation");
if (flowm_selected_relation == 1)
feat_set_string(v->features, "print_info_relation", "Word");
if (flowm_selected_relation == 2)
feat_set_string(v->features, "print_info_relation", "Segment");
/* Do the synthesis */
previous_wave = flite_text_to_wave(s,v);
ns = cst_wave_num_samples(previous_wave);
cst_free(s);
audio_flush(fl_ad);
audio_close(fl_ad);
fl_ad = NULL;
return ns;
}
int main(int argc, char **argv)
{
cst_wave *nw, *all;
cst_val *files;
const cst_val *w;
cst_val *wavelist;
cst_features *args;
int i,j;
float ntime;
int stime;
const char *nwfile;
args = new_features();
files =
cst_args(argv,argc,
"usage: combine_waves OPTIONS\n"
"Combine waves into single waveform\n"
"-o <string> Output waveform\n"
"-f <int> Input sample rate (for raw input)\n"
"-itype <string> Input type, raw or headered\n"
"-wavelist <string> File containing times and wave filenames\n",
args);
wavelist = get_wavelist(get_param_string(args,"-wavelist","-"));
if (wavelist == 0)
return -1;
all = new_wave();
for (w = wavelist; w; w = val_cdr(w))
{
ntime = decode_time(val_string(val_car(w)));
nwfile = val_string(val_car(val_cdr(w)));
nw = new_wave();
if (cst_wave_load_riff(nw,nwfile) != CST_OK_FORMAT)
{
fprintf(stderr,
"combine_waves: can't read file or wrong format \"%s\"\n",
nwfile);
continue;
}
stime = ntime * nw->sample_rate;
cst_wave_resize(all,stime+nw->num_samples,1);
for (i=0,j=stime; i<nw->num_samples; i++,j++)
{
/* this will cause overflows */
all->samples[j] += nw->samples[i];
}
delete_wave(nw);
}
cst_wave_save_riff(all,get_param_string(args,"-o","-"));
return 0;
}
void test_resize(void)
{
cst_wave *w = new_wave();
cst_wave_resize(w, 200, 2);
TEST_CHECK(w->num_samples == 200);
TEST_CHECK(w->num_channels = 2);
TEST_CHECK(cst_wave_resize(NULL, 200, 2) < 0);
delete_wave(w);
}
void test_create(void)
{
cst_wave *w = NULL;
w = new_wave();
TEST_CHECK(w != NULL);
TEST_CHECK(w->num_samples == 0);
TEST_CHECK(w->type == NULL);
delete_wave(w);
}
void test_copy(void)
{
cst_voice *v = NULL;
cst_wave *w1, *w2;
common_init();
v = mimic_voice_select(A_VOICE);
w1 = mimic_text_to_wave("Hello", v);
w2 = copy_wave(w1);
TEST_CHECK(w1 != w2);
TEST_CHECK(w1->num_samples == w2->num_samples);
TEST_CHECK(w1->sample_rate == w2->sample_rate);
TEST_CHECK(w1->num_channels == w2->num_channels);
TEST_CHECK(w1->samples != w2->samples);
TEST_CHECK(w1->samples[0] == w2->samples[0]);
TEST_CHECK(w1->samples[10] == w2->samples[10]);
TEST_CHECK(w1->samples[20] == w2->samples[20]);
delete_wave(w1);
delete_wave(w2);
}
void test_rescale(void)
{
int i;
cst_wave *w = new_wave();
cst_wave_resize(w, 10, 2);
for (i = 0; i < 10; i++)
w->samples[i] = 10 + i;
cst_wave_rescale(w, 65536 * 2); //scale x2
for (i = 0; i < 10; i++)
TEST_CHECK(w->samples[i] == (10 + i) * 2);
delete_wave(w);
}
void test_concat(void)
{
int original_len;
cst_wave *w1, *w2;
cst_voice *v;
common_init();
v = mimic_voice_select(A_VOICE);
w1 = mimic_text_to_wave("Hello", v);
w2 = mimic_text_to_wave("There", v);
original_len = w1->num_samples;
concat_wave(w1, w2);
TEST_CHECK(w1->num_samples == w2->num_samples + original_len);
delete_wave(w1);
delete_wave(w2);
w1 = mimic_text_to_wave("Hello", v);
w2 = mimic_text_to_wave("There", v);
w2->sample_rate *= 2; // create sample rate mismatch
concat_wave(w1, w2);
TEST_CHECK(w1->num_samples != w2->num_samples + original_len);
delete_wave(w1);
delete_wave(w2);
w1 = mimic_text_to_wave("Hello", v);
w2 = mimic_text_to_wave("There", v);
w2->num_channels *= 2; // create channel number mismatch
concat_wave(w1, w2);
TEST_CHECK(w1->num_samples != w2->num_samples + original_len);
delete_wave(w1);
delete_wave(w2);
}
void next_tweet(struct sockaddr_in* server, int chan)
{
char buffer[BUFSIZE+1];
memset(buffer, BUFSIZE, 0);
receive_tweet(buffer, server);
printf("%s\n", buffer);
if (si.w[chan])
delete_wave(si.w[chan]);
si.w[chan] = flite_text_to_wave(buffer, voice);
si.done[chan] = 0;
si.pos[chan] = 0;
si.cur_delay[chan] = 0.;
si.rate_delay[chan] = 44100. / (double)si.w[chan]->sample_rate;
/* play some tweets faster, some slower */
si.rate_delay[chan] *= (.75 + .5 * (double)(rand() % 256) / 255.);
}
void flowm_terminate()
{
#ifdef DOTTS
int i;
for (i=0; VoxDefs[i].name; i++)
{
(VoxDefs[i].urv)(VoxDefs[i].v); /* unregister voice */
}
#endif
if (previous_wave)
{
delete_wave(previous_wave);
previous_wave = NULL;
}
return;
}
float flite_ssml_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_tokenstream *ts;
int fp;
cst_wave *w;
float d;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for ssml reading\n",
filename);
return 1;
}
fp = get_param_int(voice->features,"file_start_position",0);
if (fp > 0)
ts_set_stream_pos(ts,fp);
/* If its a file to write to, create and save an empty wave file */
/* as we are going to incrementally append to it */
if (!cst_streq(outtype,"play") &&
!cst_streq(outtype,"none") &&
!cst_streq(outtype,"stream"))
{
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
d = flite_ssml_to_speech_ts(ts,voice,outtype);
ts_close(ts);
return d;
}
void at_exit()
{
Pa_Terminate();
int i;
if (si.w != NULL)
{
for (i = 0; i < si.channel_count; ++i)
delete_wave(si.w[i]);
free(si.w);
si.w = NULL;
}
if (voice) {
delete_voice(voice);
voice = NULL;
}
if (sock >= 0)
close(sock);
}
int flowm_say_file(TCHAR *tfilename)
{
int rc = 0;
char *filename;
cst_voice *v;
if (previous_wave)
{ /* This is really tidy up from Play -- but might say space */
delete_wave(previous_wave);
previous_wave = NULL;
}
if (fl_ad)
{
MessageBoxW(0,L"audio fd still open",L"SayFile",0);
audio_close(fl_ad);
fl_ad = NULL;
}
v = VoxDefs[flowm_selected_voice].v;
/* Where we want to start from */
feat_set_int(v->features, "file_start_position", flowm_file_pos);
/* Only do print_info in play mode */
feat_remove(v->features,"print_info_relation");
filename = cst_wstr2cstr(tfilename);
rc = flite_file_to_speech(filename, v, "stream");
cst_free(filename);
audio_flush(fl_ad);
audio_close(fl_ad);
fl_ad = NULL;
return rc;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float d, durs = 0;
int num_tokens;
cst_breakfunc breakfunc = default_utt_break;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
/* If its a file to write to delete it as we're going to */
/* incrementally append to it */
if (!cst_streq(outtype,"play") && !cst_streq(outtype,"none"))
{
cst_wave *w;
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt */
d = flite_tokens_to_speech(utt,voice,outtype);
utt = NULL;
if (d < 0)
goto out;
durs += d;
if (ts_eof(ts))
goto out;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
item_set_int(t,"file_pos",ts->file_pos);
item_set_int(t,"line_number",ts->line_number);
}
out:
delete_utterance(utt);
ts_close(ts);
return durs;
}
int main(int argc, char **argv)
{
cst_wave *w;
int port;
char *server;
char *encoding;
int i,iw;
port = CST_AUDIO_DEFAULT_PORT;
server = CST_AUDIO_DEFAULT_SERVER;
encoding = CST_AUDIO_DEFAULT_ENCODING;
if (argc == 1)
{
play_client_main_usage();
return 1;
}
if ((cst_streq("-h",argv[1])) ||
(cst_streq("-help",argv[1])) ||
(cst_streq("--help",argv[1])))
{
play_client_main_usage();
return 1;
}
iw = 1;
if (cst_streq("-s",argv[iw]))
{
if (argc < iw+1)
{
fprintf(stderr,"ERROR: no servername given\n");
play_client_main_usage();
return 1;
}
server = argv[iw+1];
iw+=2;
}
if (cst_streq("-p",argv[iw]))
{
if (argc < iw+1)
{
fprintf(stderr,"ERROR: no port given\n");
play_client_main_usage();
return 1;
}
port = atoi(argv[iw+1]);
iw+=2;
}
if (cst_streq("-e",argv[iw]))
{
if (argc < iw+1)
{
fprintf(stderr,"ERROR: no encoding given\n");
play_client_main_usage();
return 1;
}
encoding = argv[iw+1];
iw+=2;
}
for (i=iw; i<argc; i++)
{
w = new_wave();
if (cst_wave_load_riff(w,argv[i]) != CST_OK_FORMAT)
{
fprintf(stderr,
"play_wave: can't read file or wrong format \"%s\"\n",
argv[i]);
continue;
}
play_wave_client(w,server,port,encoding);
delete_wave(w);
}
return 0;
}
float flite_file_to_speech(const char *filename,
cst_voice *voice,
const char *outtype)
{
cst_utterance *utt;
cst_tokenstream *ts;
const char *token;
cst_item *t;
cst_relation *tokrel;
float durs = 0;
int num_tokens;
cst_wave *w;
cst_breakfunc breakfunc = default_utt_break;
cst_uttfunc utt_user_callback = 0;
int fp;
if ((ts = ts_open(filename,
get_param_string(voice->features,"text_whitespace",NULL),
get_param_string(voice->features,"text_singlecharsymbols",NULL),
get_param_string(voice->features,"text_prepunctuation",NULL),
get_param_string(voice->features,"text_postpunctuation",NULL)))
== NULL)
{
cst_errmsg("failed to open file \"%s\" for reading\n",
filename);
return 1;
}
fp = get_param_int(voice->features,"file_start_position",0);
if (fp > 0)
ts_set_stream_pos(ts,fp);
if (feat_present(voice->features,"utt_break"))
breakfunc = val_breakfunc(feat_val(voice->features,"utt_break"));
if (feat_present(voice->features,"utt_user_callback"))
utt_user_callback = val_uttfunc(feat_val(voice->features,"utt_user_callback"));
/* If its a file to write to, create and save an empty wave file */
/* as we are going to incrementally append to it */
if (!cst_streq(outtype,"play") &&
!cst_streq(outtype,"none") &&
!cst_streq(outtype,"stream"))
{
w = new_wave();
cst_wave_resize(w,0,1);
cst_wave_set_sample_rate(w,16000);
cst_wave_save_riff(w,outtype); /* an empty wave */
delete_wave(w);
}
num_tokens = 0;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
while (!ts_eof(ts) || num_tokens > 0)
{
token = ts_get(ts);
if ((cst_strlen(token) == 0) ||
(num_tokens > 500) || /* need an upper bound */
(relation_head(tokrel) &&
breakfunc(ts,token,tokrel)))
{
/* An end of utt, so synthesize it */
if (utt_user_callback)
utt = (utt_user_callback)(utt);
if (utt)
{
utt = flite_do_synth(utt,voice,utt_synth_tokens);
durs += flite_process_output(utt,outtype,TRUE);
delete_utterance(utt); utt = NULL;
}
else
break;
if (ts_eof(ts)) break;
utt = new_utterance();
tokrel = utt_relation_create(utt, "Token");
num_tokens = 0;
}
num_tokens++;
t = relation_append(tokrel, NULL);
item_set_string(t,"name",token);
item_set_string(t,"whitespace",ts->whitespace);
item_set_string(t,"prepunctuation",ts->prepunctuation);
item_set_string(t,"punc",ts->postpunctuation);
/* Mark it at the beginning of the token */
item_set_int(t,"file_pos",
ts->file_pos-(1+ /* as we are already on the next char */
cst_strlen(token)+
cst_strlen(ts->prepunctuation)+
cst_strlen(ts->postpunctuation)));
item_set_int(t,"line_number",ts->line_number);
}
delete_utterance(utt);
ts_close(ts);
return durs;
}
/*--------------------------------------------------------------------
* flite_synth : synthesize current text-buffer
*--------------------------------------------------------------------*/
void flite_synth(t_flite *x) {
cst_wave *wave;
int i,vecsize;
t_garray *a;
t_float *vec;
# ifdef FLITE_DEBUG
post("flite: got message 'synth'");
# endif
// -- sanity checks
if (!(a = (t_garray *)pd_findbyclass(x->x_arrayname, garray_class))) {
pd_error(x,"flite: no such array '%s'", x->x_arrayname->s_name);
return;
}
if (!x->textbuf) {
pd_error(x,"flite: attempt to synthesize empty text-buffer!");
return;
}
# ifdef FLITE_DEBUG
post("flite: flite_text_to_wave()");
# endif
wave = flite_text_to_wave(x->textbuf,voice);
if (!wave) {
pd_error(x,"flite: synthesis failed for text '%s'", x->textbuf);
return;
}
// -- resample
# ifdef FLITE_DEBUG
post("flite: cst_wave_resample()");
# endif
cst_wave_resample(wave,sys_getsr());
// -- resize & write to our array
# ifdef FLITE_DEBUG
post("flite: garray_resize(%d)", wave->num_samples);
# endif
garray_resize(a, wave->num_samples);
if (!garray_getfloatarray(a, &vecsize, &vec))
pd_error(x,"flite: bad template for write to array '%s'", x->x_arrayname->s_name);
# ifdef FLITE_DEBUG
post("flite: ->write to garray loop<-");
# endif
for (i = 0; i < wave->num_samples; i++) {
*vec++ = wave->samples[i]/32767.0;
}
// -- outlet synth-done-bang
outlet_bang(x->x_obj.ob_outlet);
// -- cleanup
delete_wave(wave);
// -- redraw
garray_redraw(a);
}
void test_delete(void)
{
delete_wave(NULL);
}
void *_flite_speak(void *nothing)
{
AudioTrack track;
cst_wave *wav;
unsigned int pos;
char *buf;
int bytes;
int ret;
log_msg(OTTS_LOG_DEBUG, "flite: speaking thread starting.......\n");
set_speaking_thread_parameters();
while (1) {
sem_wait(flite_semaphore);
log_msg(OTTS_LOG_INFO, "Semaphore on\n");
flite_stopped = 0;
flite_speaking = 1;
if (opentts_audio_set_volume(module_audio_id, flite_volume) < 0) {
log_msg(OTTS_LOG_ERR,
"Can't set volume. audio not initialized?\n");
continue;
}
/* TODO: free(buf) */
buf =
(char *)g_malloc((FliteMaxChunkLength + 1) * sizeof(char));
pos = 0;
module_report_event_begin();
while (1) {
if (flite_stopped) {
log_msg(OTTS_LOG_INFO,
"Stop in child, terminating");
flite_speaking = 0;
module_report_event_stop();
break;
}
bytes =
module_get_message_part(*flite_message, buf, &pos,
FliteMaxChunkLength,
FliteDelimiters);
if (bytes < 0) {
log_msg(OTTS_LOG_DEBUG, "End of message");
flite_speaking = 0;
module_report_event_end();
break;
}
buf[bytes] = 0;
log_msg(OTTS_LOG_DEBUG,
"Returned %d bytes from get_part\n", bytes);
log_msg(OTTS_LOG_NOTICE, "Text to synthesize is '%s'\n",
buf);
if (flite_pause_requested && (current_index_mark != -1)) {
log_msg(OTTS_LOG_INFO,
"Pause requested in parent, position %d\n",
current_index_mark);
flite_pause_requested = 0;
flite_position = current_index_mark;
break;
}
if (bytes > 0) {
log_msg(OTTS_LOG_DEBUG, "Speaking in child...");
log_msg(OTTS_LOG_NOTICE,
"Trying to synthesize text");
wav = flite_text_to_wave(buf, flite_voice);
if (wav == NULL) {
log_msg(OTTS_LOG_NOTICE,
"Stop in child, terminating");
flite_speaking = 0;
module_report_event_stop();
break;
}
track.num_samples = wav->num_samples;
track.num_channels = wav->num_channels;
track.sample_rate = wav->sample_rate;
track.bits = 16;
track.samples = wav->samples;
flite_strip_silence(&track);
log_msg(OTTS_LOG_INFO, "Got %d samples",
track.num_samples);
if (track.samples != NULL) {
if (flite_stopped) {
log_msg(OTTS_LOG_NOTICE,
"Stop in child, terminating");
flite_speaking = 0;
module_report_event_stop();
delete_wave(wav);
break;
}
log_msg(OTTS_LOG_INFO,
"Playing part of the message");
switch (module_audio_id->format) {
case SPD_AUDIO_LE:
ret =
opentts_audio_play
(module_audio_id, track,
SPD_AUDIO_LE);
break;
case SPD_AUDIO_BE:
ret =
opentts_audio_play
(module_audio_id, track,
SPD_AUDIO_BE);
break;
}
if (ret < 0)
log_msg(OTTS_LOG_WARN,
"ERROR: spd_audio failed to play the track");
if (flite_stopped) {
log_msg(OTTS_LOG_NOTICE,
"Stop in child, terminating (s)");
flite_speaking = 0;
module_report_event_stop();
delete_wave(wav);
break;
}
}
delete_wave(wav);
} else if (bytes == -1) {
log_msg(OTTS_LOG_INFO,
"End of data in speaking thread");
flite_speaking = 0;
module_report_event_end();
break;
} else {
flite_speaking = 0;
module_report_event_end();
break;
}
if (flite_stopped) {
log_msg(OTTS_LOG_NOTICE,
"Stop in child, terminating");
flite_speaking = 0;
module_report_event_stop();
break;
}
}
flite_stopped = 0;
g_free(buf);
}
flite_speaking = 0;
log_msg(OTTS_LOG_NOTICE, "flite: speaking thread ended.......\n");
pthread_exit(NULL);
}
