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;
}
cst_voice *register___VOICENAME__(const char *voxdir)
{
cst_voice *v = new_voice();
v->name = "__NICKNAME__";
/* Sets up language specific parameters in the __VOICENAME__. */
usenglish_init(v);
/* Things that weren't filled in already. */
feat_set_string(v->features,"name","__VOICENAME__");
/* Lexicon */
cmu_lex_init();
feat_set(v->features,"lexicon",lexicon_val(&cmu_lex));
/* Waveform synthesis */
feat_set(v->features,"wave_synth_func",uttfunc_val(&clunits_synth));
feat_set(v->features,"clunit_db",clunit_db_val(&__VOICENAME___db));
feat_set_int(v->features,"sample_rate",__VOICENAME___db.sts->sample_rate);
feat_set_string(v->features,"join_type","simple_join");
feat_set_string(v->features,"resynth_type","fixed");
/* Unit selection */
__VOICENAME___db.unit_name_func = __VOICENAME___unit_name;
__VOICENAME___clunits = v;
return __VOICENAME___clunits;
}
cst_voice *register___VOICENAME__(const char *voxdir)
{
cst_voice *v;
cst_lexicon *lex;
if (__VOICENAME___clunits)
return __VOICENAME___clunits; /* Already registered */
v = new_voice();
v->name = "__NICKNAME__";
/* Sets up language specific parameters in the __VOICENAME__. */
__FLITELANG___init(v);
/* Things that weren't filled in already. */
flite_feat_set_string(v->features,"name","__VOICENAME__");
/* Duration model */
flite_feat_set(v->features,"dur_cart",cart_val(&__VOICENAME___dur_cart));
flite_feat_set(v->features,"dur_stats",dur_stats_val((dur_stats *)__VOICENAME___dur_stats));
/* Lexicon */
lex = __FLITELEX___init();
flite_feat_set(v->features,"lexicon",lexicon_val(lex));
flite_feat_set(v->features,"postlex_func",uttfunc_val(lex->postlex));
/* Waveform synthesis */
flite_feat_set(v->features,"wave_synth_func",uttfunc_val(&clunits_synth));
flite_feat_set(v->features,"clunit_db",clunit_db_val(&__VOICENAME___db));
flite_feat_set_int(v->features,"sample_rate",__VOICENAME___db.sts->sample_rate);
flite_feat_set_string(v->features,"join_type","simple_join");
flite_feat_set_string(v->features,"resynth_type","fixed");
if ((voxdir != NULL) &&
(__VOICENAME___db.sts->sts == NULL) &&
(__VOICENAME___db.sts->sts_paged == NULL) &&
(__VOICENAME___db.sts->frames == NULL))
flite_mmap_clunit_voxdata(voxdir,v);
/* Unit selection */
__VOICENAME___db.unit_name_func = __VOICENAME___unit_name;
__VOICENAME___clunits = v;
return __VOICENAME___clunits;
}
cst_voice *register_cmu_us_kal(const char *voxdir)
{
cst_voice *v;
cst_lexicon *lex;
if (cmu_us_kal_diphone)
return cmu_us_kal_diphone; /* Already registered */
v = new_voice();
v->name = "kal";
/* Set up basic values for synthesizing with this voice */
usenglish_init(v);
#ifndef FLITE_PLUS_HTS_ENGINE
flite_feat_set_string(v->features,"name","cmu_us_kal_diphone");
#endif /* !FLITE_PLUS_HTS_ENGINE */
/* Lexicon */
lex = cmu_lex_init();
flite_feat_set(v->features,"lexicon",lexicon_val(lex));
#ifndef FLITE_PLUS_HTS_ENGINE
/* Intonation */
flite_feat_set_float(v->features,"int_f0_target_mean",95.0);
flite_feat_set_float(v->features,"int_f0_target_stddev",11.0);
flite_feat_set_float(v->features,"duration_stretch",1.1);
#endif /* !FLITE_PLUS_HTS_ENGINE */
/* Post lexical rules */
flite_feat_set(v->features,"postlex_func",uttfunc_val(&cmu_us_kal_postlex));
#ifndef FLITE_PLUS_HTS_ENGINE
/* Waveform synthesis: diphone_synth */
flite_feat_set(v->features,"wave_synth_func",uttfunc_val(&diphone_synth));
flite_feat_set(v->features,"diphone_db",diphone_db_val(&cmu_us_kal_db));
flite_feat_set_int(v->features,"sample_rate",cmu_us_kal_db.sts->sample_rate);
flite_feat_set_string(v->features,"resynth_type","fixed");
flite_feat_set_string(v->features,"join_type","modified_lpc");
#endif /* !FLITE_PLUS_HTS_ENGINE */
cmu_us_kal_diphone = v;
return cmu_us_kal_diphone;
}
void usenglish_init(cst_voice *v)
{
us_text_init();
/* utterance break function */
feat_set(v->features,"utt_break",breakfunc_val(&default_utt_break));
/* Phoneset */
feat_set(v->features,"phoneset",phoneset_val(&us_phoneset));
feat_set_string(v->features,"silence",us_phoneset.silence);
/* Text analyser */
feat_set_string(v->features,"text_whitespace",us_english_whitespace);
feat_set_string(v->features,"text_postpunctuation",us_english_punctuation);
feat_set_string(v->features,"text_prepunctuation",
us_english_prepunctuation);
feat_set_string(v->features,"text_singlecharsymbols",
us_english_singlecharsymbols);
feat_set(v->features,"tokentowords_func",itemfunc_val(&us_tokentowords));
/* Phrasing */
feat_set(v->features,"phrasing_cart",cart_val(&us_phrasing_cart));
/* Intonation */
feat_set(v->features,"int_cart_accents",cart_val(&us_int_accent_cart));
feat_set(v->features,"int_cart_tones",cart_val(&us_int_tone_cart));
#ifndef FLITE_PLUS_HTS_ENGINE
/* Duration */
feat_set(v->features,"dur_cart",cart_val(&us_durz_cart));
feat_set(v->features,"dur_stats",dur_stats_val((dur_stats *)us_dur_stats));
/* f0 model */
feat_set(v->features,"f0_model_func",uttfunc_val(&us_f0_model));
#endif /* !FLITE_PLUS_HTS_ENGINE */
/* Post lexical rules */
feat_set(v->features,"postlex_func",uttfunc_val(&us_postlex));
us_ff_register(v->ffunctions);
}
void russian_init(cst_utterance *u)
{
feat_set(u->features,"phoneset",phoneset_val(&ru_phoneset));
feat_set_string(u->features,"silence",ru_phoneset.silence);
feat_set(u->features,"textanalysis_func",uttfunc_val(&russian_textanalysis));
feat_set(u->features,"lexical_insertion_func",uttfunc_val(&russian_lexical_insertion));
feat_set(u->features,"phrasing_func",uttfunc_val(&russian_phrasify));
feat_set(u->features,"pause_insertion_func",uttfunc_val(&russian_pause_insertion));
feat_set(u->features,"intonation_func",uttfunc_val(&do_nothing));
feat_set(u->features,"postlex_func",uttfunc_val(russian_postlex_function));
feat_set(u->features,"duration_model_func",uttfunc_val(do_nothing));
feat_set(u->features,"f0_model_func",uttfunc_val(do_nothing));
feat_set_string(u->features,"no_segment_duration_model","1");
feat_set_string(u->features,"no_f0_target_model","1");
feat_set(u->features,"wave_synth_func",uttfunc_val(&hts_synth));
ru_ff_register(u->ffunctions);
}
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;
}
void flowm_init()
{
#ifdef DOTTS
int i;
cst_audio_streaming_info *asi;
flite_init(); /* Initialize flite interface */
for (i=0; VoxDefs[i].name; i++)
{
VoxDefs[i].v = (VoxDefs[i].rv)(NULL); /* register voice */
/* Set up call back function for low level audio streaming */
/* This way it plays the waveform as it synthesizes it */
/* This is necessary for the slower (CG) voices */
asi = new_audio_streaming_info();
asi->asc = flowm_audio_stream_chunk;
asi->min_buffsize = VoxDefs[i].min_buffsize;
feat_set(VoxDefs[i].v->features,
"streaming_info",
audio_streaming_info_val(asi));
/* Set up call back function for sending what tokens are being */
/* synthesized and for keeping track of the current position in */
/* the file */
feat_set(VoxDefs[i].v->features,
"utt_user_callback",
uttfunc_val(flowm_utt_callback));
/* For outputing results of a relation (only used in play) */
feat_set(VoxDefs[i].v->features,
"post_synth_hook_func",
uttfunc_val(flowm_print_relation_callback));
}
#endif
return;
}
cst_voice *register_cmu_us_rms(const char *voxdir)
{
cst_voice *vox;
cst_lexicon *lex;
if (cmu_us_rms_cg)
return cmu_us_rms_cg; /* Already registered */
vox = new_voice();
vox->name = "rms";
/* Sets up language specific parameters in the cmu_us_rms. */
usenglish_init(vox);
/* Things that weren't filled in already. */
flite_feat_set_string(vox->features,"name","cmu_us_rms");
/* Lexicon */
lex = cmu_lex_init();
flite_feat_set(vox->features,"lexicon",lexicon_val(lex));
flite_feat_set(vox->features,"postlex_func",uttfunc_val(lex->postlex));
/* No standard segment durations are needed as its done at the */
/* HMM state level */
flite_feat_set_string(vox->features,"no_segment_duration_model","1");
flite_feat_set_string(vox->features,"no_f0_target_model","1");
/* Waveform synthesis */
flite_feat_set(vox->features,"wave_synth_func",uttfunc_val(&cg_synth));
flite_feat_set(vox->features,"cg_db",cg_db_val(&cmu_us_rms_cg_db));
flite_feat_set_int(vox->features,"sample_rate",cmu_us_rms_cg_db.sample_rate);
cmu_us_rms_cg = vox;
return cmu_us_rms_cg;
}
cst_voice *register___VOICENAME__(const char *voxdir)
{
cst_voice *vox;
cst_lexicon *lex;
if (__VOICENAME___cg)
return __VOICENAME___cg; /* Already registered */
vox = new_voice();
vox->name = "__NICKNAME__";
/* Sets up language specific parameters in the __VOICENAME__. */
__MIMICLANG___init(vox);
/* Things that weren't filled in already. */
mimic_feat_set_string(vox->features,"name","__VOICENAME__");
/* Lexicon */
lex = __MIMICLEX___init();
mimic_feat_set(vox->features,"lexicon",lexicon_val(lex));
mimic_feat_set(vox->features,"postlex_func",uttfunc_val(lex->postlex));
/* No standard segment durations are needed as its done at the */
/* HMM state level */
mimic_feat_set_string(vox->features,"no_segment_duration_model","1");
mimic_feat_set_string(vox->features,"no_f0_target_model","1");
/* Waveform synthesis */
mimic_feat_set(vox->features,"wave_synth_func",uttfunc_val(&cg_synth));
mimic_feat_set(vox->features,"cg_db",cg_db_val(&__VOICENAME___cg_db));
mimic_feat_set_int(vox->features,"sample_rate",__VOICENAME___cg_db.sample_rate);
__VOICENAME___cg = vox;
return __VOICENAME___cg;
}
cst_voice *register_cmu_time_awb(const char *voxdir)
{
cst_voice *v = new_voice();
v->name = "awb_time";
/* Sets up language specific parameters in the voice. */
usenglish_init(v);
/* Things that weren't filled in already. */
flite_feat_set_string(v->features,"name","cmu_time_awb");
/* Lexicon */
cmu_time_awb_lex.name = "cmu_time_awb";
cmu_time_awb_lex.num_entries = cmu_time_awb_num_entries;
cmu_time_awb_lex.num_bytes = cmu_time_awb_num_bytes;
cmu_time_awb_lex.data = (unsigned char*) cmu_time_awb_lex_data;
cmu_time_awb_lex.phone_table = (char**) cmu_time_awb_lex_phone_table;
cmu_time_awb_lex.syl_boundary = cmu_syl_boundary;
cmu_time_awb_lex.lts_rule_set = NULL;
cmu_time_awb_lex.phone_hufftable = cmu_time_awb_lex_phones_huff_table;
cmu_time_awb_lex.entry_hufftable = cmu_time_awb_lex_entries_huff_table;
flite_feat_set(v->features,"lexicon",lexicon_val(&cmu_time_awb_lex));
/* Waveform synthesis */
flite_feat_set(v->features,"wave_synth_func",uttfunc_val(&clunits_synth));
flite_feat_set(v->features,"clunit_db",clunit_db_val(&cmu_time_awb_db));
flite_feat_set_int(v->features,"sample_rate",cmu_time_awb_db.sts->sample_rate);
flite_feat_set_string(v->features,"join_type","simple_join");
flite_feat_set_string(v->features,"resynth_type","fixed");
/* Unit selection */
cmu_time_awb_db.unit_name_func = cmu_time_awb_unit_name;
cmu_time_awb_ldom = v;
return cmu_time_awb_ldom;
}
int main(int argc, char **argv)
{
struct timeval tv;
cst_voice *v;
const char *filename;
const char *outtype;
int i;
float durs;
double time_start, time_end;
int flite_verbose, flite_loop, flite_bench;
int explicit_filename, explicit_text, explicit_phones;
#define ITER_MAX 3
int bench_iter = 0;
cst_features *extra_feats;
filename = 0;
outtype = "play"; /* default is to play */
flite_verbose = FALSE;
flite_loop = FALSE;
flite_bench = FALSE;
explicit_text = explicit_filename = explicit_phones = FALSE;
extra_feats = new_features();
flite_init();
for (i=1; i<argc; i++)
{
if (cst_streq(argv[i],"--version"))
{
flite_version();
return 1;
}
else if (cst_streq(argv[i],"-h") ||
cst_streq(argv[i],"--help") ||
cst_streq(argv[i],"-?"))
flite_usage();
else if (cst_streq(argv[i],"-v"))
flite_verbose = TRUE;
else if (cst_streq(argv[i],"-l"))
flite_loop = TRUE;
else if (cst_streq(argv[i],"-b"))
{
flite_bench = TRUE;
break; /* ignore other arguments */
}
else if ((cst_streq(argv[i],"-o")) && (i+1 < argc))
{
outtype = argv[i+1];
i++;
}
else if (cst_streq(argv[i],"-f") && (i+1 < argc))
{
filename = argv[i+1];
explicit_filename = TRUE;
i++;
}
else if (cst_streq(argv[i],"-pw"))
{
feat_set_string(extra_feats,"print_info_relation","Word");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ps"))
{
feat_set_string(extra_feats,"print_info_relation","Segment");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-pr") && (i+1 < argc))
{
feat_set_string(extra_feats,"print_info_relation",argv[i+1]);
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
i++;
}
else if ((cst_streq(argv[i],"-set") || cst_streq(argv[i],"-s"))
&& (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],0);
i++;
}
else if (cst_streq(argv[i],"--seti") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"int");
i++;
}
else if (cst_streq(argv[i],"--setf") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"float");
i++;
}
else if (cst_streq(argv[i],"--sets") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"string");
i++;
}
else if (cst_streq(argv[i],"-p") && (i+1 < argc))
{
filename = argv[i+1];
explicit_phones = TRUE;
i++;
}
else if (cst_streq(argv[i],"-t") && (i+1 < argc))
{
filename = argv[i+1];
explicit_text = TRUE;
i++;
}
else if (filename)
outtype = argv[i];
else
filename = argv[i];
}
if (filename == NULL) filename = "-"; /* stdin */
v = REGISTER_VOX(NULL);
feat_copy_into(extra_feats,v->features);
durs = 0.0;
if (flite_bench)
{
outtype = "none";
filename = "A whole joy was reaping, but they've gone south, you should fetch azure mike.";
explicit_text = TRUE;
}
loop:
gettimeofday(&tv,NULL);
time_start = (double)(tv.tv_sec)+(((double)tv.tv_usec)/1000000.0);
if (explicit_phones)
durs = flite_phones_to_speech(filename,v,outtype);
else if ((strchr(filename,' ') && !explicit_filename) || explicit_text)
durs = flite_text_to_speech(filename,v,outtype);
else
durs = flite_file_to_speech(filename,v,outtype);
gettimeofday(&tv,NULL);
time_end = ((double)(tv.tv_sec))+((double)tv.tv_usec/1000000.0);
if (flite_verbose || (flite_bench && bench_iter == ITER_MAX))
printf("times faster than real-time: %f\n(%f seconds of speech synthesized in %f)\n",
durs/(float)(time_end-time_start),
durs,
(float)(time_end-time_start));
if (flite_loop || (flite_bench && bench_iter++ < ITER_MAX))
goto loop;
delete_features(extra_feats);
UNREGISTER_VOX(v);
return 0;
}
cst_voice *cst_cg_load_voice(const char *filename,
const cst_lang *lang_table)
{
cst_voice *vox;
cst_lexicon *lex = NULL;
int i, end_of_features;
const char *language;
const char *xname;
cst_cg_db *cg_db;
char* fname;
char* fval;
cst_file vd;
vd = cst_fopen(filename,CST_OPEN_READ | CST_OPEN_BINARY);
if (vd == NULL)
{
cst_errmsg("Error load voice: can't open file %s\n",filename);
return NULL;
}
if (cst_cg_read_header(vd) != 0)
{
cst_errmsg("Error load voice: %s does not have expected header\n",filename);
cst_fclose(vd);
return NULL;
}
vox = new_voice();
/* Read voice features from the external file */
/* Read until the feature is "end_of_features" */
fname="";
end_of_features = 0;
while (end_of_features == 0)
{
cst_read_voice_feature(vd,&fname, &fval);
if (cst_streq(fname,"end_of_features"))
end_of_features = 1;
else
{
xname = feat_own_string(vox->features,fname);
flite_feat_set_string(vox->features,xname, fval);
}
cst_free(fname);
cst_free(fval);
}
/* Load up cg_db from external file */
cg_db = cst_cg_load_db(vox,vd);
if (cg_db == NULL)
{
cst_fclose(vd);
return NULL;
}
/* Use the language feature to initialize the correct voice */
language = flite_get_param_string(vox->features, "language", "");
/* Search Lang table for lang_init() and lex_init(); */
for (i=0; lang_table[i].lang; i++)
{
if (cst_streq(language,lang_table[i].lang))
{
(lang_table[i].lang_init)(vox);
lex = (lang_table[i].lex_init)();
break;
}
}
if (lex == NULL)
{ /* Language is not supported */
/* Delete allocated memory in cg_db */
cst_cg_free_db(vd,cg_db);
cst_fclose(vd);
cst_errmsg("Error load voice: lang/lex %s not supported in this binary\n",language);
return NULL;
}
/* Things that weren't filled in already. */
vox->name = cg_db->name;
flite_feat_set_string(vox->features,"name",cg_db->name);
flite_feat_set_string(vox->features,"pathname",filename);
flite_feat_set(vox->features,"lexicon",lexicon_val(lex));
flite_feat_set(vox->features,"postlex_func",uttfunc_val(lex->postlex));
/* No standard segment durations are needed as its done at the */
/* HMM state level */
flite_feat_set_string(vox->features,"no_segment_duration_model","1");
flite_feat_set_string(vox->features,"no_f0_target_model","1");
/* Waveform synthesis */
flite_feat_set(vox->features,"wave_synth_func",uttfunc_val(&cg_synth));
flite_feat_set(vox->features,"cg_db",cg_db_val(cg_db));
flite_feat_set_int(vox->features,"sample_rate",cg_db->sample_rate);
cst_fclose(vd);
return vox;
}
int main(int argc, char **argv)
{
struct timeval tv;
cst_voice *v;
const char *filename;
const char *outtype;
cst_voice *desired_voice = 0;
const char *voicedir = NULL;
int i;
float durs;
double time_start, time_end;
int flite_verbose, flite_loop, flite_bench;
int explicit_filename, explicit_text, explicit_phones, ssml_mode;
#define ITER_MAX 3
int bench_iter = 0;
cst_features *extra_feats;
const char *lex_addenda_file = NULL;
const char *voicedumpfile = NULL;
cst_audio_streaming_info *asi;
filename = 0;
outtype = "play"; /* default is to play */
flite_verbose = FALSE;
flite_loop = FALSE;
flite_bench = FALSE;
explicit_text = explicit_filename = explicit_phones = FALSE;
ssml_mode = FALSE;
extra_feats = new_features();
flite_init();
flite_add_lang("eng",usenglish_init,cmu_lex_init);
for (i=1; i<argc; i++)
{
if (cst_streq(argv[i],"--version"))
{
flite_version();
return 1;
}
else if (cst_streq(argv[i],"-h") ||
cst_streq(argv[i],"--help") ||
cst_streq(argv[i],"-?"))
flite_usage();
else if (cst_streq(argv[i],"-v"))
flite_verbose = TRUE;
else if (cst_streq(argv[i],"-lv"))
{
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
flite_voice_list_print();
exit(0);
}
else if (cst_streq(argv[i],"-l"))
flite_loop = TRUE;
else if (cst_streq(argv[i],"-b"))
{
flite_bench = TRUE;
break; /* ignore other arguments */
}
else if ((cst_streq(argv[i],"-o")) && (i+1 < argc))
{
outtype = argv[i+1];
i++;
}
else if ((cst_streq(argv[i],"-voice")) && (i+1 < argc))
{
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
desired_voice = flite_voice_select(argv[i+1]);
i++;
}
else if ((cst_streq(argv[i],"-voicedir")) && (i+1 < argc))
{
voicedir = argv[i+1];
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
i++;
}
else if ((cst_streq(argv[i],"-add_lex")) && (i+1 < argc))
{
lex_addenda_file = argv[i+1];
i++;
}
else if (cst_streq(argv[i],"-f") && (i+1 < argc))
{
filename = argv[i+1];
explicit_filename = TRUE;
i++;
}
else if (cst_streq(argv[i],"-pw"))
{
feat_set_string(extra_feats,"print_info_relation","Word");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ps"))
{
feat_set_string(extra_feats,"print_info_relation","Segment");
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
}
else if (cst_streq(argv[i],"-ssml"))
{
ssml_mode = TRUE;
}
else if (cst_streq(argv[i],"-pr") && (i+1 < argc))
{
feat_set_string(extra_feats,"print_info_relation",argv[i+1]);
feat_set(extra_feats,"post_synth_hook_func",
uttfunc_val(&print_info));
i++;
}
else if (cst_streq(argv[i],"-voicedump") && (i+1 < argc))
{
voicedumpfile = argv[i+1];
i++;
}
else if ((cst_streq(argv[i],"-set") || cst_streq(argv[i],"-s"))
&& (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],0);
i++;
}
else if (cst_streq(argv[i],"--seti") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"int");
i++;
}
else if (cst_streq(argv[i],"--setf") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"float");
i++;
}
else if (cst_streq(argv[i],"--sets") && (i+1 < argc))
{
ef_set(extra_feats,argv[i+1],"string");
i++;
}
else if (cst_streq(argv[i],"-p") && (i+1 < argc))
{
filename = argv[i+1];
explicit_phones = TRUE;
i++;
}
else if (cst_streq(argv[i],"-t") && (i+1 < argc))
{
filename = argv[i+1];
explicit_text = TRUE;
i++;
}
else if (filename)
outtype = argv[i];
else
filename = argv[i];
}
if (filename == NULL) filename = "-"; /* stdin */
if (flite_voice_list == NULL)
flite_set_voice_list(voicedir);
if (desired_voice == 0)
desired_voice = flite_voice_select(NULL);
v = desired_voice;
feat_copy_into(extra_feats,v->features);
durs = 0.0;
if (voicedumpfile != NULL)
{
flite_voice_dump(v,voicedumpfile);
exit(0);
}
if (lex_addenda_file)
flite_voice_add_lex_addenda(v,lex_addenda_file);
if (cst_streq("stream",outtype))
{
asi = new_audio_streaming_info();
asi->asc = audio_stream_chunk;
feat_set(v->features,"streaming_info",audio_streaming_info_val(asi));
}
if (flite_bench)
{
outtype = "none";
filename = "A whole joy was reaping, but they've gone south, you should fetch azure mike.";
explicit_text = TRUE;
}
loop:
gettimeofday(&tv,NULL);
time_start = (double)(tv.tv_sec)+(((double)tv.tv_usec)/1000000.0);
if (explicit_phones)
durs = flite_phones_to_speech(filename,v,outtype);
else if ((strchr(filename,' ') && !explicit_filename) || explicit_text)
{
if (ssml_mode)
durs = flite_ssml_text_to_speech(filename,v,outtype);
else
durs = flite_text_to_speech(filename,v,outtype);
}
else
{
if (ssml_mode)
durs = flite_ssml_file_to_speech(filename,v,outtype);
else
durs = flite_file_to_speech(filename,v,outtype);
}
gettimeofday(&tv,NULL);
time_end = ((double)(tv.tv_sec))+((double)tv.tv_usec/1000000.0);
if (flite_verbose || (flite_bench && bench_iter == ITER_MAX))
printf("times faster than real-time: %f\n(%f seconds of speech synthesized in %f)\n",
durs/(float)(time_end-time_start),
durs,
(float)(time_end-time_start));
if (flite_loop || (flite_bench && bench_iter++ < ITER_MAX))
goto loop;
delete_features(extra_feats);
delete_val(flite_voice_list); flite_voice_list=0;
/* cst_alloc_debug_summary(); */
return 0;
}
