cst_voice* register_cmu_us_no_wave( const char* voxdir )
{
cst_voice* v = new_voice();
cst_lexicon* lex;
v->name = "no_wave_voice";
/* Set up basic values for synthesizing with this voice */
usenglish_init( v );
feat_set_string( v->features, "name", "cmu_us_no_wave" );
/* Lexicon */
lex = cmu_lex_init();
feat_set( v->features, "lexicon", lexicon_val( lex ) );
/* Intonation */
feat_set_float( v->features, "int_f0_target_mean", 95.0 );
feat_set_float( v->features, "int_f0_target_stddev", 11.0 );
feat_set_float( v->features, "duration_stretch", 1.1 );
/* Post lexical rules */
feat_set( v->features, "postlex_func", uttfunc_val( lex->postlex ) );
/* Waveform synthesis: diphone_synth */
feat_set( v->features, "wave_synth_func", uttfunc_val( &no_wave_synth ) );
cmu_us_no_wave = v;
return cmu_us_no_wave;
}
cst_voice *register_cmu_us_kal16(const char *voxdir)
{
cst_voice *v = new_voice();
usenglish_init(v);
/* Set up basic values for synthesizing with this voice */
feat_set_string(v->features,"name","cmu_us_kal_diphone");
/* Lexicon */
cmu_lex_init();
feat_set(v->features,"lexicon",lexicon_val(&cmu_lex));
/* Intonation */
feat_set_float(v->features,"int_f0_target_mean",105.0);
feat_set_float(v->features,"int_f0_target_stddev",14.0);
/* Post lexical rules */
feat_set(v->features,"postlex_func",uttfunc_val(&cmu_us_kal_postlex));
/* Duration */
feat_set_float(v->features,"duration_stretch",1.1);
/* Waveform synthesis: diphone_synth */
feat_set(v->features,"wave_synth_func",uttfunc_val(&diphone_synth));
feat_set(v->features,"diphone_db",diphone_db_val(&cmu_us_kal_db));
feat_set_int(v->features,"sample_rate",cmu_us_kal_db.sts->sample_rate);
feat_set_string(v->features,"resynth_type","fixed");
feat_set_string(v->features,"join_type","modified_lpc");
cmu_us_kal_diphone = v;
return cmu_us_kal_diphone;
}
static void ef_set(cst_features *f,const char *fv,const char *type)
{
/* set feature from fv (F=V), guesses type if not explicit type given */
const char *val;
char *feat;
if ((val = strchr(fv,'=')) == 0)
{
fprintf(stderr,
"flite: can't find '=' in featval \"%s\", ignoring it\n",
fv);
}
else
{
feat = cst_strdup(fv);
feat[strlen(fv)-strlen(val)] = '\0';
val = val+1;
if ((type && cst_streq("int",type)) ||
((type == 0) && (cst_regex_match(cst_rx_int,val))))
feat_set_int(f,feat,atoi(val));
else if ((type && cst_streq("float",type)) ||
((type == 0) && (cst_regex_match(cst_rx_double,val))))
feat_set_float(f,feat,atof(val));
else
feat_set_string(f,feat,val);
/* I don't free feat, because feats think featnames are const */
/* which is true except in this particular case */
}
}
static void ef_set(cst_features *f,const char *fv,const char *type)
{
/* set feature from fv (F=V), guesses type if not explicit type given */
const char *val;
char *feat;
if ((val = strchr(fv,'=')) == 0)
{
fprintf(stderr,
"flite: can't find '=' in featval \"%s\", ignoring it\n",
fv);
}
else
{
feat = cst_strdup(fv);
feat[cst_strlen(fv)-cst_strlen(val)] = '\0';
val = val+1;
if ((type && cst_streq("int",type)) ||
((type == 0) && (cst_regex_match(cst_rx_int,val))))
feat_set_int(f,feat,atoi(val));
else if ((type && cst_streq("float",type)) ||
((type == 0) && (cst_regex_match(cst_rx_double,val))))
feat_set_float(f,feat,atof(val));
else
feat_set_string(f,feat,val);
cst_free(feat);
}
}
static void flite_set_pitch(signed int pitch)
{
float f0;
assert(pitch >= OTTS_VOICE_PITCH_MIN && pitch <= OTTS_VOICE_PITCH_MAX);
f0 = (((float)pitch) * 0.8) + 100.0;
feat_set_float(flite_voice->features, "int_f0_target_mean", f0);
}
cst_val *cst_args(char **argv, int argc,
const char *description,
cst_features *args)
{
/* parses the given arguments wrt the description */
cst_features *op_types = new_features();
cst_val *files = NULL;
int i;
const char *type;
parse_description(description,op_types);
for (i=1; i<argc; i++)
{
if (argv[i][0] == '-')
{
if ((!feat_present(op_types,argv[i])) ||
(cst_streq("-h",argv[i])) ||
(cst_streq("-?",argv[i])) ||
(cst_streq("--help",argv[i])) ||
(cst_streq("-help",argv[i])))
parse_usage(argv[0],"","",description);
else
{
type = feat_string(op_types,argv[i]);
if (cst_streq("<binary>",type))
feat_set_string(args,argv[i],"true");
else
{
if (i+1 == argc)
parse_usage(argv[0],
"missing argument for ",argv[i],
description);
if (cst_streq("<int>",type))
feat_set_int(args,argv[i],atoi(argv[i+1]));
else if (cst_streq("<float>",type))
feat_set_float(args,argv[i],atof(argv[i+1]));
else if (cst_streq("<string>",type))
feat_set_string(args,argv[i],argv[i+1]);
else
parse_usage(argv[0],
"unknown arg type ",type,
description);
i++;
}
}
}
else
files = cons_val(string_val(argv[i]),files);
}
delete_features(op_types);
return val_reverse(files);
}
cst_voice *register_cmu_us_kal16(const char *voxdir)
{
cst_voice *v;
cst_lexicon *lex;
if (cmu_us_kal16_diphone)
return cmu_us_kal16_diphone; /* Already registered */
v = new_voice();
v->name = "kal16";
/* Sets up language specific parameters in the cmu_us_kal16. */
usenglish_init(v);
feat_set_string(v->features,"name","cmu_us_kal16");
feat_set_float(v->features,"int_f0_target_mean",95.0);
feat_set_float(v->features,"int_f0_target_stddev",11.0);
feat_set_float(v->features,"duration_stretch",1.1);
/* Lexicon */
lex = cmu_lex_init();
feat_set(v->features,"lexicon",lexicon_val(lex));
feat_set(v->features,"postlex_func",uttfunc_val(lex->postlex));
/* Waveform synthesis */
feat_set(v->features,"wave_synth_func",uttfunc_val(&diphone_synth));
feat_set(v->features,"diphone_db",diphone_db_val(&cmu_us_kal16_db));
feat_set_int(v->features,"sample_rate",cmu_us_kal16_db.sts->sample_rate);
/* feat_set_string(v->features,"join_type","simple_join"); */
feat_set_string(v->features,"join_type","modified_lpc");
feat_set_string(v->features,"resynth_type","fixed");
cmu_us_kal16_diphone = v;
return cmu_us_kal16_diphone;
}
static void flite_set_rate(signed int rate)
{
const float stretch_default = 1., stretch_min = 3., stretch_max =
(1 - 100. / 175.);
float stretch;
assert(rate >= OTTS_VOICE_RATE_MIN && rate <= OTTS_VOICE_RATE_MAX);
if (rate < OTTS_VOICE_RATE_DEFAULT)
stretch = stretch_min + ((float)(rate - OTTS_VOICE_RATE_MIN))
* (stretch_default - stretch_min)
/ ((float)(OTTS_VOICE_RATE_DEFAULT - OTTS_VOICE_RATE_MIN));
else
stretch =
stretch_default + (((float)(rate - OTTS_VOICE_RATE_DEFAULT))
* (stretch_max - stretch_default)
/ ((float)
(OTTS_VOICE_RATE_MAX -
OTTS_VOICE_RATE_DEFAULT)));
feat_set_float(flite_voice->features, "duration_stretch", stretch);
}
void item_set_float(const cst_item *i,const char *name,float val)
{
feat_set_float(item_feats(i),name,val);
}
void flite_feat_set_float(cst_features *f, const char *name, float v)
{
feat_set_float(f,name,v);
}