int sync_espeak_terminated_msg( uint32_t unique_identifier, void* user_data)
{//=====================================================================
ENTER("sync_espeak_terminated_msg");
int finished=0;
memset(event_list, 0, 2*sizeof(espeak_EVENT));
event_list[0].type = espeakEVENT_MSG_TERMINATED;
event_list[0].unique_identifier = unique_identifier;
event_list[0].user_data = user_data;
event_list[1].type = espeakEVENT_LIST_TERMINATED;
event_list[1].unique_identifier = unique_identifier;
event_list[1].user_data = user_data;
if (my_mode==AUDIO_OUTPUT_PLAYBACK)
{
while(1)
{
espeak_ERROR a_error = event_declare(event_list);
if (a_error != EE_BUFFER_FULL)
{
break;
}
SHOW_TIME("sync_espeak_terminated_msg > EE_BUFFER_FULL\n");
usleep(10000);
}
}
else
{
if (synth_callback)
{
finished=synth_callback(NULL,0,event_list);
}
}
return finished;
}
int sync_espeak_terminated_msg(uint32_t unique_identifier, void *user_data)
{
int finished = 0;
memset(event_list, 0, 2*sizeof(espeak_EVENT));
event_list[0].type = espeakEVENT_MSG_TERMINATED;
event_list[0].unique_identifier = unique_identifier;
event_list[0].user_data = user_data;
event_list[1].type = espeakEVENT_LIST_TERMINATED;
event_list[1].unique_identifier = unique_identifier;
event_list[1].user_data = user_data;
if (my_mode == ENOUTPUT_MODE_SPEAK_AUDIO) {
while (1) {
err = event_declare(event_list);
if (err != ENS_EVENT_BUFFER_FULL)
break;
usleep(10000);
}
} else if (synth_callback)
finished = synth_callback(NULL, 0, event_list);
return finished;
}
static espeak_ng_STATUS Synthesize(unsigned int unique_identifier, const void *text, int flags)
{
// Fill the buffer with output sound
int length;
int finished = 0;
int count_buffers = 0;
if ((outbuf == NULL) || (event_list == NULL))
return ENS_NOT_INITIALIZED;
option_ssml = flags & espeakSSML;
option_phoneme_input = flags & espeakPHONEMES;
option_endpause = flags & espeakENDPAUSE;
count_samples = 0;
espeak_ng_STATUS status;
if (translator == NULL) {
status = espeak_ng_SetVoiceByName("en");
if (status != ENS_OK)
return status;
}
if (p_decoder == NULL)
p_decoder = create_text_decoder();
status = text_decoder_decode_string_multibyte(p_decoder, text, translator->encoding, flags);
if (status != ENS_OK)
return status;
SpeakNextClause(0);
for (;;) {
out_ptr = outbuf;
out_end = &outbuf[outbuf_size];
event_list_ix = 0;
WavegenFill();
length = (out_ptr - outbuf)/2;
count_samples += length;
event_list[event_list_ix].type = espeakEVENT_LIST_TERMINATED; // indicates end of event list
event_list[event_list_ix].unique_identifier = unique_identifier;
event_list[event_list_ix].user_data = my_user_data;
count_buffers++;
if ((my_mode & ENOUTPUT_MODE_SPEAK_AUDIO) == ENOUTPUT_MODE_SPEAK_AUDIO) {
finished = create_events((short *)outbuf, length, event_list);
if (finished < 0)
return ENS_AUDIO_ERROR;
} else if (synth_callback)
finished = synth_callback((short *)outbuf, length, event_list);
if (finished) {
SpeakNextClause(2); // stop
return ENS_SPEECH_STOPPED;
}
if (Generate(phoneme_list, &n_phoneme_list, 1) == 0) {
if (WcmdqUsed() == 0) {
// don't process the next clause until the previous clause has finished generating speech.
// This ensures that <audio> tag (which causes end-of-clause) is at a sound buffer boundary
event_list[0].type = espeakEVENT_LIST_TERMINATED;
event_list[0].unique_identifier = my_unique_identifier;
event_list[0].user_data = my_user_data;
if (SpeakNextClause(1) == 0) {
finished = 0;
if ((my_mode & ENOUTPUT_MODE_SPEAK_AUDIO) == ENOUTPUT_MODE_SPEAK_AUDIO) {
if (dispatch_audio(NULL, 0, NULL) < 0)
return ENS_AUDIO_ERROR;
} else if (synth_callback)
finished = synth_callback(NULL, 0, event_list); // NULL buffer ptr indicates end of data
if (finished) {
SpeakNextClause(2); // stop
return ENS_SPEECH_STOPPED;
}
return ENS_OK;
}
}
}
}
}
static int dispatch_audio(short *outbuf, int length, espeak_EVENT *event)
{
int a_wave_can_be_played = 1;
#ifdef USE_ASYNC
if ((my_mode & ENOUTPUT_MODE_SYNCHRONOUS) == 0)
a_wave_can_be_played = fifo_is_command_enabled();
#endif
switch ((int)my_mode)
{
case ENOUTPUT_MODE_SPEAK_AUDIO:
case ENOUTPUT_MODE_SPEAK_AUDIO | ENOUTPUT_MODE_SYNCHRONOUS:
{
int event_type = 0;
if (event)
event_type = event->type;
if (event_type == espeakEVENT_SAMPLERATE) {
voice_samplerate = event->id.number;
if (out_samplerate != voice_samplerate) {
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if (out_samplerate != 0) {
// sound was previously open with a different sample rate
audio_object_close(my_audio);
out_samplerate = 0;
#ifdef HAVE_SLEEP
sleep(1);
#endif
}
#endif
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
int error = audio_object_open(my_audio, AUDIO_OBJECT_FORMAT_S16LE, voice_samplerate, 1);
if (error != 0) {
fprintf(stderr, "error: %s\n", audio_object_strerror(my_audio, error));
err = ENS_AUDIO_ERROR;
return -1;
}
#endif
out_samplerate = voice_samplerate;
#ifdef USE_ASYNC
if ((my_mode & ENOUTPUT_MODE_SYNCHRONOUS) == 0)
event_init();
#endif
}
}
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if (out_samplerate == 0) {
int error = audio_object_open(my_audio, AUDIO_OBJECT_FORMAT_S16LE, voice_samplerate, 1);
if (error != 0) {
fprintf(stderr, "error: %s\n", audio_object_strerror(my_audio, error));
err = ENS_AUDIO_ERROR;
return -1;
}
out_samplerate = voice_samplerate;
}
#endif
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if (outbuf && length && a_wave_can_be_played) {
int error = audio_object_write(my_audio, (char *)outbuf, 2*length);
if (error != 0)
fprintf(stderr, "error: %s\n", audio_object_strerror(my_audio, error));
}
#endif
#ifdef USE_ASYNC
while (event && a_wave_can_be_played) {
// TBD: some event are filtered here but some insight might be given
// TBD: in synthesise.cpp for avoiding to create WORDs with size=0.
// TBD: For example sentence "or ALT)." returns three words
// "or", "ALT" and "".
// TBD: the last one has its size=0.
if ((event->type == espeakEVENT_WORD) && (event->length == 0))
break;
if ((my_mode & ENOUTPUT_MODE_SYNCHRONOUS) == 0) {
err = event_declare(event);
if (err != ENS_EVENT_BUFFER_FULL)
break;
usleep(10000);
a_wave_can_be_played = fifo_is_command_enabled();
} else
break;
}
#endif
}
break;
case 0:
if (synth_callback)
synth_callback(outbuf, length, event);
break;
}
return a_wave_can_be_played == 0; // 1 = stop synthesis, -1 = error
}
static int dispatch_audio(short* outbuf, int length, espeak_EVENT* event)
{//======================================================================
ENTER("dispatch_audio");
int a_wave_can_be_played = fifo_is_command_enabled();
#ifdef DEBUG_ENABLED
SHOW("*** dispatch_audio > uid=%d, [write=%p (%d bytes)], sample=%d, a_wave_can_be_played = %d\n",
(event) ? event->unique_identifier : 0, wave_test_get_write_buffer(), 2*length,
(event) ? event->sample : 0,
a_wave_can_be_played);
#endif
switch(my_mode)
{
case AUDIO_OUTPUT_PLAYBACK:
{
int event_type=0;
if(event)
{
event_type = event->type;
}
if(event_type == espeakEVENT_SAMPLERATE)
{
voice_samplerate = event->id.number;
if(out_samplerate != voice_samplerate)
{
if(out_samplerate != 0)
{
// sound was previously open with a different sample rate
wave_close(my_audio);
sleep(1);
}
out_samplerate = voice_samplerate;
if(!wave_init(voice_samplerate))
{
err = EE_INTERNAL_ERROR;
return(-1);
}
wave_set_callback_is_output_enabled( fifo_is_command_enabled);
my_audio = wave_open("alsa");
event_init();
}
}
if (outbuf && length && a_wave_can_be_played)
{
wave_write (my_audio, (char*)outbuf, 2*length);
}
while(a_wave_can_be_played) {
// TBD: some event are filtered here but some insight might be given
// TBD: in synthesise.cpp for avoiding to create WORDs with size=0.
// TBD: For example sentence "or ALT)." returns three words
// "or", "ALT" and "".
// TBD: the last one has its size=0.
if (event && (event->type == espeakEVENT_WORD) && (event->length==0))
{
break;
}
espeak_ERROR a_error = event_declare(event);
if (a_error != EE_BUFFER_FULL)
{
break;
}
SHOW_TIME("dispatch_audio > EE_BUFFER_FULL\n");
usleep(10000);
a_wave_can_be_played = fifo_is_command_enabled();
}
}
break;
case AUDIO_OUTPUT_RETRIEVAL:
if (synth_callback)
{
synth_callback(outbuf, length, event);
}
break;
case AUDIO_OUTPUT_SYNCHRONOUS:
case AUDIO_OUTPUT_SYNCH_PLAYBACK:
break;
}
if (!a_wave_can_be_played)
{
SHOW_TIME("dispatch_audio > synth must be stopped!\n");
}
SHOW_TIME("LEAVE dispatch_audio\n");
return (a_wave_can_be_played==0); // 1 = stop synthesis, -1 = error
}
static espeak_ERROR Synthesize(unsigned int unique_identifier, const void *text, int flags)
{//========================================================================================
// Fill the buffer with output sound
int length;
int finished = 0;
int count_buffers = 0;
#ifdef USE_ASYNC
uint32_t a_write_pos=0;
#endif
#ifdef DEBUG_ENABLED
ENTER("Synthesize");
if (text)
{
SHOW("Synthesize > uid=%d, flags=%d, >>>text=%s<<<\n", unique_identifier, flags, text);
}
#endif
if((outbuf==NULL) || (event_list==NULL))
return(EE_INTERNAL_ERROR); // espeak_Initialize() has not been called
option_multibyte = flags & 7;
option_ssml = flags & espeakSSML;
option_phoneme_input = flags & espeakPHONEMES;
option_endpause = flags & espeakENDPAUSE;
count_samples = 0;
#ifdef USE_ASYNC
if(my_mode == AUDIO_OUTPUT_PLAYBACK)
{
a_write_pos = wave_get_write_position(my_audio);
}
#endif
if(translator == NULL)
{
SetVoiceByName("default");
}
SpeakNextClause(NULL,text,0);
if(my_mode == AUDIO_OUTPUT_SYNCH_PLAYBACK)
{
for(;;)
{
#ifdef PLATFORM_WINDOWS
Sleep(300); // 0.3s
#else
#ifdef USE_NANOSLEEP
struct timespec period;
struct timespec remaining;
period.tv_sec = 0;
period.tv_nsec = 300000000; // 0.3 sec
nanosleep(&period,&remaining);
#else
sleep(1);
#endif
#endif
if(SynthOnTimer() != 0)
break;
}
return(EE_OK);
}
for(;;)
{
#ifdef DEBUG_ENABLED
SHOW("Synthesize > %s\n","for (next)");
#endif
out_ptr = outbuf;
out_end = &outbuf[outbuf_size];
event_list_ix = 0;
WavegenFill(0);
length = (out_ptr - outbuf)/2;
count_samples += length;
event_list[event_list_ix].type = espeakEVENT_LIST_TERMINATED; // indicates end of event list
event_list[event_list_ix].unique_identifier = my_unique_identifier;
event_list[event_list_ix].user_data = my_user_data;
count_buffers++;
if (my_mode==AUDIO_OUTPUT_PLAYBACK)
{
#ifdef USE_ASYNC
finished = create_events((short *)outbuf, length, event_list, a_write_pos);
if(finished < 0)
return EE_INTERNAL_ERROR;
length = 0; // the wave data are played once.
#endif
}
else
{
finished = synth_callback((short *)outbuf, length, event_list);
}
if(finished)
{
SpeakNextClause(NULL,0,2); // stop
break;
}
if(Generate(phoneme_list,&n_phoneme_list,1)==0)
{
if(WcmdqUsed() == 0)
{
// don't process the next clause until the previous clause has finished generating speech.
// This ensures that <audio> tag (which causes end-of-clause) is at a sound buffer boundary
event_list[0].type = espeakEVENT_LIST_TERMINATED;
event_list[0].unique_identifier = my_unique_identifier;
event_list[0].user_data = my_user_data;
if(SpeakNextClause(NULL,NULL,1)==0)
{
#ifdef USE_ASYNC
if (my_mode==AUDIO_OUTPUT_PLAYBACK)
{
if(dispatch_audio(NULL, 0, NULL) < 0) // TBD: test case
return err = EE_INTERNAL_ERROR;
}
else
{
synth_callback(NULL, 0, event_list); // NULL buffer ptr indicates end of data
}
#else
synth_callback(NULL, 0, event_list); // NULL buffer ptr indicates end of data
#endif
break;
}
}
}
}
return(EE_OK);
} // end of Synthesize
