float flite_phones_to_speech(const char *text,
cst_voice *voice,
const char *outtype)
{
cst_utterance *u;
float dur;
u = flite_synth_phones(text,voice);
dur = flite_process_output(u,outtype,FALSE);
delete_utterance(u);
return dur;
}
static float flite_ssml_to_speech_ts(cst_tokenstream *ts,
cst_voice *voice,
const char *outtype)
{
cst_features *ssml_feats, *ssml_word_feats;
cst_features *attributes;
const char *token;
char *tag;
cst_utterance *utt;
cst_relation *tokrel;
int num_tokens;
cst_breakfunc breakfunc = default_utt_break;
cst_uttfunc utt_user_callback = 0;
float durs = 0.0;
cst_item *t;
ssml_feats = new_features();
ssml_word_feats = new_features();
set_charclasses(ts,
" \t\n\r",
ssml_singlecharsymbols_general,
get_param_string(voice->features,"text_prepunctuation",""),
get_param_string(voice->features,"text_postpunctuation","")
);
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"));
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_streq("<",token))
{ /* A tag */
tag = cst_upcase(ts_get(ts));
if (cst_streq("/",tag)) /* an end tag */
{
tag = cst_upcase(ts_get(ts));
attributes = ssml_get_attributes(ts);
feat_set_string(attributes,"_type","end");
}
else
attributes = ssml_get_attributes(ts);
utt = ssml_apply_tag(tag,attributes,utt,ssml_word_feats);
cst_free(tag);
}
else if (cst_streq("&",token))
{ /* an escape sequence */
/* skip to ; and insert value in rawdata */
}
else
{
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);
return durs;
}
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;
}
float flite_text_to_speech_phenome( const char* text,
cst_voice* voice,
const char* outtype,
void* pStream )
{
cst_utterance* u;
float dur;
float end_last = 0;
float end_current = 0;
float dur_current = 0;
float dur_sum = 0;
//feat_set_float( voice->features, "duration_stretch", 1 );
u = flite_synth_text( text, voice );
cst_item* s;
string sRet;
int nPhoneme = 0;
for ( s = relation_head( utt_relation( u, "Segment" ) ); s; s = item_next( s ) )
{
SPhenomeTiming ps;
string sPhoneme = item_feat_string( s, "name" );
sRet += sPhoneme;
end_current = item_feat_float( s, "end" );
dur_current = end_current - end_last;
//if ( !( nPhoneme == 0 && sPhoneme == "pau" ) )
//{
dur_sum += dur_current;
//}
ps.fWeight = 1;
/* If its a vowel and is stressed output stress value */
if ( ( cst_streq( "+", ffeature_string( s, "ph_vc" ) ) ) &&
( cst_streq( "1", ffeature_string( s, "R:SylStructure.parent.stress" ) ) ) )
{
sRet += "1";
ps.fWeight = 1.3;
}
sRet += " ";
if ( pStream )
{
// fade into each other
ps.sName = sPhoneme;
ps.fStart = end_current - dur_current;
ps.fEnd = end_current;
ps.fDuration = dur_current;
( ( CryMT::queue<SPhenomeTiming>* )pStream )->push( ps );
}
end_last = end_current;
++nPhoneme;
}
dur = flite_process_output( u, outtype, FALSE );
delete_utterance( u );
return dur;
}
