static void cepstral_text_param_tts(switch_speech_handle_t *sh, char *param, const char *val)
{
cepstral_t *cepstral;
cepstral = sh->private_info;
assert(cepstral != NULL);
if (!strcasecmp(param, "voice")) {
const char *voice_name = val;
if (!strcasecmp(voice_name, "next")) {
if ((cepstral->voice = swift_port_find_next_voice(cepstral->port))) {
if (SWIFT_FAILED(swift_port_set_voice(cepstral->port, cepstral->voice))) {
cepstral->done = cepstral->done_gen = 1;
return;
}
voice_name = swift_voice_get_attribute(cepstral->voice, "name");
} else {
voice_name = NULL;
}
} else {
if (voice_name && SWIFT_FAILED(swift_port_set_voice_by_name(cepstral->port, voice_name))) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Invalid voice %s!\n", voice_name);
voice_name = NULL;
}
}
if (!voice_name) {
/* Find the first voice on the system */
if ((cepstral->voice = swift_port_find_first_voice(cepstral->port, NULL, NULL)) == NULL) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to find any voices!\n");
cepstral->done = cepstral->done_gen = 1;
return;
}
/* Set the voice found by find_first_voice() as the port's current voice */
if (SWIFT_FAILED(swift_port_set_voice(cepstral->port, cepstral->voice))) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to set voice.\n");
cepstral->done = cepstral->done_gen = 1;
return;
}
voice_name = swift_voice_get_attribute(cepstral->voice, "name");
}
if (voice_name) {
switch_copy_string(sh->voice, voice_name, sizeof(sh->voice));
}
return;
}
swift_port_set_param_string(cepstral->port, param, val, NULL);
}
/** Swift engine callback */
static swift_result_t mrcp_swift_write_audio(swift_event *event, swift_event_t type, void *udata)
{
void *buf;
int len;
mrcp_swift_channel_t *synth_channel = udata;
swift_event_t rv = SWIFT_SUCCESS;
if(type & SWIFT_EVENT_END) {
apt_log(APT_LOG_MARK,APT_PRIO_DEBUG,"Swift Engine: Write End-of-Speech Event");
mpf_buffer_event_write(synth_channel->audio_buffer,MEDIA_FRAME_TYPE_EVENT);
return rv;
}
rv = swift_event_get_audio(event, &buf, &len);
if(!SWIFT_FAILED(rv)) {
#if 0
/* Get the event times */
float time_start, time_len;
swift_event_get_times(event, &time_start, &time_len);
apt_log(APT_LOG_MARK,APT_PRIO_DEBUG,"Swift Engine: Write Audio [%d | %0.4f | %0.4f]",len, time_start, time_len);
#endif
mpf_buffer_audio_write(synth_channel->audio_buffer,buf,len);
}
return rv;
}
/** Set Swift port param */
static apt_bool_t mrcp_swift_channel_param_set(mrcp_swift_channel_t *synth_channel, const char *name, swift_val *val)
{
swift_result_t res;
if(SWIFT_FAILED(res = swift_port_set_param(synth_channel->port,name,val,synth_channel->tts_stream)) ) {
const char *error_string = swift_strerror(res);
apt_log(APT_LOG_MARK,APT_PRIO_DEBUG,"Swift Param %s: %s",name,error_string);
return FALSE;
}
return TRUE;
}
/* This callback caches the audio in the buffer */
static swift_result_t write_audio(swift_event * event, swift_event_t type, void *udata)
{
cepstral_t *cepstral;
swift_event_t rv = SWIFT_SUCCESS;
void *buf = NULL;
int len = 0, i = 0;
cepstral = udata;
assert(cepstral != NULL);
if (!cepstral->port || cepstral->done || cepstral->done_gen) {
return SWIFT_UNKNOWN_ERROR;
}
/* Only proceed when we have success */
if (!SWIFT_FAILED((rv = swift_event_get_audio(event, &buf, &len)))) {
while (!cepstral->done) {
switch_mutex_lock(cepstral->audio_lock);
if (switch_buffer_write(cepstral->audio_buffer, buf, len) > 0) {
switch_mutex_unlock(cepstral->audio_lock);
break;
}
switch_mutex_unlock(cepstral->audio_lock);
if (!cepstral->done) {
for (i = 0; i < 10; i++) {
switch_yield(10000);
if (cepstral->done) {
break;
}
}
}
}
} else {
cepstral->done = 1;
}
if (cepstral->done) {
rv = SWIFT_UNKNOWN_ERROR;
}
return rv;
}
/** Set voice matching specified criteria */
static apt_bool_t mrcp_swift_channel_voice_set(mrcp_swift_channel_t *synth_channel, mrcp_message_t *message)
{
mrcp_synth_header_t *synth_header = mrcp_resource_header_get(message);
char search_criteria[1024];
int offset = 0;
swift_voice *voice;
swift_result_t res;
if(!synth_header) {
/* no params to set */
return TRUE;
}
if(mrcp_resource_header_property_check(message,SYNTHESIZER_HEADER_VOICE_NAME) == TRUE) {
offset += search_criteria_delimiter_add(search_criteria+offset,sizeof(search_criteria)-offset,(offset == 0));
offset += apr_snprintf(search_criteria+offset,sizeof(search_criteria)-offset,"speaker/name=%s",synth_header->voice_param.name.buf);
}
if(mrcp_resource_header_property_check(message,SYNTHESIZER_HEADER_VOICE_GENDER) == TRUE) {
switch(synth_header->voice_param.gender) {
case VOICE_GENDER_MALE:
offset += search_criteria_delimiter_add(search_criteria+offset,sizeof(search_criteria)-offset,offset == 0);
offset += apr_snprintf(search_criteria+offset,sizeof(search_criteria)-offset,"speaker/gender=male");
break;
case VOICE_GENDER_FEMALE:
offset += search_criteria_delimiter_add(search_criteria+offset,sizeof(search_criteria)-offset,offset == 0);
offset += apr_snprintf(search_criteria+offset,sizeof(search_criteria)-offset,"speaker/gender=female");
break;
default:
break;
}
}
if(mrcp_resource_header_property_check(message,SYNTHESIZER_HEADER_VOICE_AGE) == TRUE) {
offset += search_criteria_delimiter_add(search_criteria+offset,sizeof(search_criteria)-offset,offset == 0);
offset += apr_snprintf(search_criteria+offset,sizeof(search_criteria)-offset,"speaker/age=%d",synth_header->voice_param.age);
}
if(mrcp_resource_header_property_check(message,SYNTHESIZER_HEADER_SPEECH_LANGUAGE) == TRUE) {
const char *swift_lang_name = NULL;
if(swift_speech_language_table) {
swift_lang_name = apr_table_get(swift_speech_language_table,synth_header->speech_language.buf);
}
if(!swift_lang_name) {
swift_lang_name = synth_header->speech_language.buf;
}
offset += search_criteria_delimiter_add(search_criteria+offset,sizeof(search_criteria)-offset,offset == 0);
offset += apr_snprintf(search_criteria+offset,sizeof(search_criteria)-offset,"language/name=%s",swift_lang_name);
}
if(offset > 0) {
apt_log(APT_LOG_MARK,APT_PRIO_INFO,"Find Voices Matching the Criteria [%s]",search_criteria);
if((voice = swift_port_find_first_voice(synth_channel->port,search_criteria,NULL)) == NULL) {
apt_log(APT_LOG_MARK,APT_PRIO_INFO,"No Swift Voice Available Matching the Criteria [%s]",search_criteria);
voice = swift_port_find_first_voice(synth_channel->port,NULL,NULL);
}
if(SWIFT_FAILED(res = swift_port_set_voice(synth_channel->port,voice)) ) {
const char *error_string = swift_strerror(res);
apt_log(APT_LOG_MARK,APT_PRIO_INFO,error_string);
return FALSE;
}
}
return TRUE;
}
static int app_exec(struct ast_channel *chan, void *data)
{
int res = 0, argc = 0, max_digits = 0, timeout = 0, alreadyran = 0, old_writeformat = 0;
int ms, len, availatend;
char *argv[3], *parse = NULL, *text = NULL, *rc = NULL;
char tmp_exten[2], results[20];
struct ast_module_user *u;
struct ast_frame *f;
struct timeval next;
struct stuff *ps;
struct myframe {
struct ast_frame f;
unsigned char offset[AST_FRIENDLY_OFFSET];
unsigned char frdata[framesize];
} myf;
swift_engine *engine;
swift_port *port = NULL;
swift_voice *voice;
swift_params *params;
swift_result_t sresult;
swift_background_t tts_stream;
unsigned int event_mask;
memset(results, 0 ,20);
memset(tmp_exten, 0, 2);
memset(argv, 0, 3);
parse = ast_strdupa(data);
u = ast_module_user_add(chan);
argc = ast_app_separate_args(parse, ',', argv, 3);
text = argv[0];
if (!ast_strlen_zero(argv[1])) {
timeout = strtol(argv[1], NULL, 0);
}
if (!ast_strlen_zero(argv[2])) {
max_digits = strtol(argv[2], NULL, 0);
}
if (ast_strlen_zero(text)) {
ast_log(LOG_WARNING, "%s requires text to speak!\n", app);
return -1;
}
if (!ast_strlen_zero(text)) {
ast_log(LOG_DEBUG, "Text to Speak : %s\n", text);
}
if (timeout > 0) {
ast_log(LOG_DEBUG, "Timeout : %d\n", timeout);
}
if (max_digits > 0) {
ast_log(LOG_DEBUG, "Max Digits : %d\n", max_digits);
}
ps = malloc(sizeof(struct stuff));
swift_init_stuff(ps);
/* Setup synthesis */
if ((engine = swift_engine_open(NULL)) == NULL) {
ast_log(LOG_ERROR, "Failed to open Swift Engine.\n");
goto exception;
}
params = swift_params_new(NULL);
swift_params_set_string(params, "audio/encoding", "ulaw");
swift_params_set_string(params, "audio/sampling-rate", "8000");
swift_params_set_string(params, "audio/output-format", "raw");
swift_params_set_string(params, "tts/text-encoding", "utf-8");
/* Additional swift parameters
*
* swift_params_set_float(params, "speech/pitch/shift", 1.0);
* swift_params_set_int(params, "speech/rate", 150);
* swift_params_set_int(params, "audio/volume", 110);
* swift_params_set_int(params, "audio/deadair", 0);
*/
if ((port = swift_port_open(engine, params)) == NULL) {
ast_log(LOG_ERROR, "Failed to open Swift Port.\n");
goto exception;
}
if ((voice = swift_port_set_voice_by_name(port, cfg_voice)) == NULL) {
ast_log(LOG_ERROR, "Failed to set voice.\n");
goto exception;
}
event_mask = SWIFT_EVENT_AUDIO | SWIFT_EVENT_END;
swift_port_set_callback(port, &swift_cb, event_mask, ps);
if (SWIFT_FAILED(swift_port_speak_text(port, text, 0, NULL, &tts_stream, NULL))) {
ast_log(LOG_ERROR, "Failed to speak.\n");
goto exception;
}
if (chan->_state != AST_STATE_UP) {
ast_answer(chan);
}
ast_stopstream(chan);
old_writeformat = chan->writeformat;
if (ast_set_write_format(chan, AST_FORMAT_ULAW) < 0) {
ast_log(LOG_WARNING, "Unable to set write format.\n");
goto exception;
}
res = 0;
/* Wait 100ms first for synthesis to start crankin'; if that's not
* enough the
*/
next = ast_tvadd(ast_tvnow(), ast_tv(0, 100000));
while (swift_generator_running(ps)) {
ms = ast_tvdiff_ms(next, ast_tvnow());
if (ms <= 0) {
if (swift_bytes_available(ps) > 0) {
ASTOBJ_WRLOCK(ps);
len = fmin(framesize, ps->qc);
availatend = cfg_buffer_size - (ps->pq_r - ps->q);
if (len > availatend) {
/* read #1: to end of q buf */
memcpy(myf.frdata, ps->pq_r, availatend);
ps->qc -= availatend;
/* read #2: reset to start of q buf and get rest */
ps->pq_r = ps->q;
memcpy(myf.frdata + availatend, ps->pq_r, len - availatend);
ps->qc -= len - availatend;
ps->pq_r += len - availatend;
} else {
ast_log(LOG_DEBUG, "Easy read; %d bytes and %d at end, %d free\n", len, availatend, cfg_buffer_size - ps->qc);
memcpy(myf.frdata, ps->pq_r, len);
ps->qc -= len;
ps->pq_r += len;
}
myf.f.frametype = AST_FRAME_VOICE;
myf.f.subclass = AST_FORMAT_ULAW;
myf.f.datalen = len;
myf.f.samples = len;
myf.f.data.ptr = myf.frdata;
myf.f.mallocd = 0;
myf.f.offset = AST_FRIENDLY_OFFSET;
myf.f.src = __PRETTY_FUNCTION__;
myf.f.delivery.tv_sec = 0;
myf.f.delivery.tv_usec = 0;
if (ast_write(chan, &myf.f) < 0) {
ast_log(LOG_DEBUG, "ast_write failed\n");
}
ast_log(LOG_DEBUG, "wrote a frame of %d\n", len);
if (ps->qc < 0) {
ast_log(LOG_DEBUG, "queue claims to contain negative bytes. Huh? qc < 0\n");
}
ASTOBJ_UNLOCK(ps);
next = ast_tvadd(next, ast_samp2tv(myf.f.samples, samplerate));
} else {
next = ast_tvadd(next, ast_samp2tv(framesize / 2, samplerate));
ast_log(LOG_DEBUG, "Whoops, writer starved for audio\n");
}
} else {
ms = ast_waitfor(chan, ms);
if (ms < 0) {
ast_log(LOG_DEBUG, "Hangup detected\n");
res = -1;
ASTOBJ_WRLOCK(ps);
ps->immediate_exit = 1;
ASTOBJ_UNLOCK(ps);
} else if (ms) {
f = ast_read(chan);
if (!f) {
ast_log(LOG_DEBUG, "Null frame == hangup() detected\n");
res = -1;
ASTOBJ_WRLOCK(ps);
ps->immediate_exit = 1;
ASTOBJ_UNLOCK(ps);
} else if (f->frametype == AST_FRAME_DTMF && timeout > 0 && max_digits > 0) {
char originDTMF = f->subclass;
alreadyran = 1;
res = 0;
ASTOBJ_WRLOCK(ps);
ps->immediate_exit = 1;
ASTOBJ_UNLOCK(ps);
if (max_digits > 1) {
rc = listen_for_dtmf(chan, timeout, max_digits - 1);
}
if (rc) {
sprintf(results, "%c%s", originDTMF, rc);
} else {
sprintf(results, "%c", originDTMF);
}
ast_log(LOG_NOTICE, "DTMF = %s\n", results);
pbx_builtin_setvar_helper(chan, "SWIFT_DTMF", results);
}
ast_frfree(f);
}
}
ASTOBJ_RDLOCK(ps);
if (ps->immediate_exit && !ps->generating_done) {
if (SWIFT_FAILED(sresult = swift_port_stop(port, tts_stream, SWIFT_EVENT_NOW))) {
ast_log(LOG_NOTICE, "Early top of swift port failed\n");
}
}
ASTOBJ_UNLOCK(ps);
}
if (alreadyran == 0 && timeout > 0 && max_digits > 0) {
rc = listen_for_dtmf(chan, timeout, max_digits);
if (rc != NULL) {
sprintf(results, "%s", rc);
ast_log(LOG_NOTICE, "DTMF = %s\n", results);
pbx_builtin_setvar_helper(chan, "SWIFT_DTMF", results);
}
}
if (max_digits >= 1 && results != NULL) {
if (cfg_goto_exten) {
ast_log(LOG_NOTICE, "GoTo(%s|%s|%d) : ", chan->context, results, 1);
if (ast_exists_extension (chan, chan->context, results, 1, chan->cid.cid_num)) {
ast_log(LOG_NOTICE, "OK\n");
ast_copy_string(chan->exten, results, sizeof(chan->exten) - 1);
chan->priority = 0;
} else {
ast_log(LOG_NOTICE, "FAILED\n");
}
}
}
exception:
if (port != NULL) {
swift_port_close(port);
}
if (engine != NULL) {
swift_engine_close(engine);
}
if (ps && ps->q) {
ast_free(ps->q);
ps->q = NULL;
}
if (ps) {
ast_free(ps);
ps = NULL;
}
if (!res && old_writeformat) {
ast_set_write_format(chan, old_writeformat);
}
ast_module_user_remove(u);
return res;
}
static swift_result_t swift_cb(swift_event *event, swift_event_t type, void *udata)
{
void *buf;
int len, spacefree;
unsigned long sleepfor;
swift_event_t rv = SWIFT_SUCCESS;
struct stuff *ps = udata;
if (type == SWIFT_EVENT_AUDIO) {
rv = swift_event_get_audio(event, &buf, &len);
if (!SWIFT_FAILED(rv) && len > 0) {
ast_log(LOG_DEBUG, "audio callback\n");
ASTOBJ_WRLOCK(ps);
/* Sleep while waiting for some queue space to become available */
while (len + ps->qc > cfg_buffer_size && !ps->immediate_exit) {
/* Each byte is 125us of time, so assume queue space will become available
at that rate and guess when we'll have enough space available.
+ another (125 usec/sample * framesize samples) (1 frame) for fudge */
sleepfor = ((unsigned long)(len - (cfg_buffer_size - ps->qc)) * 125UL) + (125UL * (unsigned long)framesize);
/* ast_log(LOG_DEBUG, "generator: %d bytes to write but only %d space avail, sleeping %ldus\n", len, cfg_buffer_size - ps->qc, sleepfor); */
ASTOBJ_UNLOCK(ps);
usleep(sleepfor);
ASTOBJ_WRLOCK(ps);
}
if (ps->immediate_exit) {
ASTOBJ_UNLOCK(ps);
return SWIFT_SUCCESS;
}
spacefree = cfg_buffer_size - ((uintptr_t) ps->pq_w - (uintptr_t)ps->q);
if (len > spacefree) {
ast_log(LOG_DEBUG, "audio fancy write; %d bytes but only %d avail to end %d totalavail\n", len, spacefree, cfg_buffer_size - ps->qc);
/* write #1 to end of mem */
memcpy(ps->pq_w, buf, spacefree);
ps->pq_w = ps->q;
ps->qc += spacefree;
/* write #2 and beg of mem */
memcpy(ps->pq_w, buf + spacefree, len - spacefree);
ps->pq_w += len - spacefree;
ps->qc += len - spacefree;
} else {
ast_log(LOG_DEBUG, "audio easy write, %d avail to end %d totalavail\n", spacefree, cfg_buffer_size - ps->qc);
memcpy(ps->pq_w, buf, len);
ps->pq_w += len;
ps->qc += len;
}
ASTOBJ_UNLOCK(ps);
} else {
ast_log(LOG_DEBUG, "got audio callback but get_audio call failed\n");
}
} else if (type == SWIFT_EVENT_END) {
ast_log(LOG_DEBUG, "got END callback; done generating audio\n");
ASTOBJ_WRLOCK(ps);
ps->generating_done = 1;
ASTOBJ_UNLOCK(ps);
} else {
ast_log(LOG_DEBUG, "UNKNOWN callback\n");
}
return rv;
}
static switch_status_t cepstral_speech_open(switch_speech_handle_t *sh, const char *voice_name, int rate, switch_speech_flag_t *flags)
{
cepstral_t *cepstral = switch_core_alloc(sh->memory_pool, sizeof(*cepstral));
char srate[25];
if (!cepstral) {
return SWITCH_STATUS_MEMERR;
}
if (switch_buffer_create_dynamic(&cepstral->audio_buffer, MY_BLOCK_SIZE, MY_BUF_LEN, 0) != SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_CRIT, "Write Buffer Failed!\n");
return SWITCH_STATUS_MEMERR;
}
switch_mutex_init(&cepstral->audio_lock, SWITCH_MUTEX_NESTED, sh->memory_pool);
cepstral->params = swift_params_new(NULL);
swift_params_set_string(cepstral->params, "audio/encoding", "pcm16");
switch_snprintf(srate, sizeof(srate), "%d", rate);
swift_params_set_string(cepstral->params, "audio/sampling-rate", srate);
/* Open a Swift Port through which to make TTS calls */
if (!(cepstral->port = swift_port_open(engine, cepstral->params))) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to open Swift Port.\n");
goto all_done;
}
if (voice_name && SWIFT_FAILED(swift_port_set_voice_by_name(cepstral->port, voice_name))) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Invalid voice %s!\n", voice_name);
voice_name = NULL;
}
if (zstr(voice_name)) {
/* Find the first voice on the system */
if ((cepstral->voice = swift_port_find_first_voice(cepstral->port, NULL, NULL)) == NULL) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to find any voices!\n");
goto all_done;
}
/* Set the voice found by find_first_voice() as the port's current voice */
if (SWIFT_FAILED(swift_port_set_voice(cepstral->port, cepstral->voice))) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to set voice.\n");
goto all_done;
}
voice_name = (char *) swift_voice_get_attribute(cepstral->voice, "name");
}
if (voice_name) {
switch_copy_string(sh->voice, voice_name, sizeof(sh->voice));
}
swift_port_set_callback(cepstral->port, &write_audio, SWIFT_EVENT_AUDIO, cepstral);
sh->private_info = cepstral;
return SWITCH_STATUS_SUCCESS;
all_done:
return SWITCH_STATUS_FALSE;
}
