mahta.fetrat
/
HomoFast-eSpeak-Persian
Watch 1
Star 0
Fork 0
Code Issues 0 Pull Requests 0 Releases 0 Wiki Activity
Browse Source

Revert commit 14a2fb39cf 

This commit has been reported to cause distortions in phoneme
pitches and speeds.
master
Reece H. Dunn 9 years ago
parent
932f530a81
commit
522e7c4d51
24 changed files with 825 additions and 431 deletions
Split View Show Diff Stats
  1. 16
    8
      src/espeak-ng.c
  2. 146
    79
      src/libespeak-ng/compiledata.c
  3. 52
    32
      src/libespeak-ng/compiledict.c
  4. 2
    1
      src/libespeak-ng/compilembrola.c
  5. 57
    29
      src/libespeak-ng/dictionary.c
  6. 5
    3
      src/libespeak-ng/ieee80.c
  7. 43
    25
      src/libespeak-ng/intonation.c
  8. 45
    21
      src/libespeak-ng/klatt.c
  9. 17
    11
      src/libespeak-ng/mbrowrap.c
  10. 45
    22
      src/libespeak-ng/numbers.c
  11. 14
    8
      src/libespeak-ng/phonemelist.c
  12. 53
    27
      src/libespeak-ng/readclause.c
  13. 20
    10
      src/libespeak-ng/setlengths.c
  14. 13
    6
      src/libespeak-ng/speak_lib.c
  15. 8
    5
      src/libespeak-ng/spect.c
  16. 18
    11
      src/libespeak-ng/synth_mbrola.c
  17. 39
    21
      src/libespeak-ng/synthdata.c
  18. 62
    30
      src/libespeak-ng/synthesize.c
  19. 14
    6
      src/libespeak-ng/tr_languages.c
  20. 34
    16
      src/libespeak-ng/translate.c
  21. 29
    15
      src/libespeak-ng/voices.c
  22. 4
    2
      src/libespeak-ng/wave_sada.c
  23. 72
    34
      src/libespeak-ng/wavegen.c
  24. 17
    9
      src/speak-ng.c

+ 16
- 8
src/espeak-ng.c View File

@@ -113,6 +113,7 @@ char wavefile[200];

void DisplayVoices(FILE *f_out, char *language)
{
int ix;
const char *p;
int len;
int count;
@@ -139,7 +140,7 @@ void DisplayVoices(FILE *f_out, char *language)

fprintf(f_out, "Pty Language Age/Gender VoiceName File Other Languages\n");

for (int ix = 0; (v = voices[ix]) != NULL; ix++) {
for (ix = 0; (v = voices[ix]) != NULL; ix++) {
count = 0;
p = v->languages;
while (*p != 0) {
@@ -173,7 +174,9 @@ void DisplayVoices(FILE *f_out, char *language)
static void Write4Bytes(FILE *f, int value)
{
// Write 4 bytes to a file, least significant first
for (int ix = 0; ix < 4; ix++) {
int ix;

for (ix = 0; ix < 4; ix++) {
fputc(value & 0xff, f);
value = value >> 8;
}
@@ -214,11 +217,13 @@ int OpenWavFile(char *path, int rate)

static void CloseWavFile()
{
unsigned int pos;

if ((f_wavfile == NULL) || (f_wavfile == stdout))
return;

fflush(f_wavfile);
unsigned int pos = ftell(f_wavfile);
pos = ftell(f_wavfile);

fseek(f_wavfile, 4, SEEK_SET);
Write4Bytes(f_wavfile, pos - 8);
@@ -232,6 +237,8 @@ static void CloseWavFile()

static int SynthCallback(short *wav, int numsamples, espeak_EVENT *events)
{
char fname[210];

if (quiet) return 0; // -q quiet mode

if (wav == NULL) {
@@ -255,7 +262,6 @@ static int SynthCallback(short *wav, int numsamples, espeak_EVENT *events)
}

if (f_wavfile == NULL) {
char fname[210];
if (samples_split > 0) {
sprintf(fname, "%s_%.2d%s", wavefile, wavefile_count+1, filetype);
if (OpenWavFile(fname, samplerate) != 0)
@@ -329,6 +335,7 @@ int main(int argc, char **argv)

int option_index = 0;
int c;
int ix;
char *optarg2;
int value;
int flag_stdin = 0;
@@ -362,6 +369,7 @@ int main(int argc, char **argv)
optind = 1;
opt_string = "";
while (optind < argc) {
int len;
char *p;

if ((c = *opt_string) == 0) {
@@ -381,10 +389,10 @@ int main(int argc, char **argv)
break; // -- means don't interpret further - as commands

opt_string = "";
for (int ix = 0;; ix++) {
for (ix = 0;; ix++) {
if (long_options[ix].name == 0)
break;
size_t len = strlen(long_options[ix].name);
len = strlen(long_options[ix].name);
if (memcmp(long_options[ix].name, p, len) == 0) {
c = long_options[ix].val;
optarg2 = NULL;
@@ -486,7 +494,7 @@ int main(int argc, char **argv)
case 0x103: // --punct
option_punctuation = 1;
if (optarg2 != NULL) {
int ix = 0;
ix = 0;
while ((ix < N_PUNCTLIST) && ((option_punctlist[ix] = optarg2[ix]) != 0)) ix++;
option_punctlist[N_PUNCTLIST-1] = 0;
option_punctuation = 2;
@@ -665,7 +673,7 @@ int main(int argc, char **argv)
}
} else {
// bulk input on stdin
int ix = 0;
ix = 0;
while (!feof(stdin)) {
p_text[ix++] = fgetc(stdin);
if (ix >= (max-1)) {

+ 146
- 79
src/libespeak-ng/compiledata.c View File

@@ -386,6 +386,7 @@ static void ReadPhondataManifest()
// Read the phondata-manifest file
FILE *f;
int n_lines = 0;
int ix;
char *p;
unsigned int value;
char buf[sizeof(path_home)+40];
@@ -401,7 +402,7 @@ static void ReadPhondataManifest()
rewind(f);

if (manifest != NULL) {
for (int ix = 0; ix < n_manifest; ix++)
for (ix = 0; ix < n_manifest; ix++)
free(manifest[ix].name);
}

@@ -442,7 +443,13 @@ static const char *KeyToMnem(keywtab_t *ktab, int type, int value)
static void DecompilePhoneme(FILE *f_out, PHONEME_TAB *ph, int compile_phoneme)
{
USHORT *pc;
int instn;
int instn_category;
int address, address2;
int data1;
int type2;
int ix;
int any;
const char *name;
char buf[120];

@@ -486,10 +493,10 @@ static void DecompilePhoneme(FILE *f_out, PHONEME_TAB *ph, int compile_phoneme)

pc = prog_buf;
while (pc < prog_out) {
int instn = *pc++;
int instn_category = (instn >> 12) & 0xf;
int data1 = instn & 0xff;
int type2 = (instn >> 8) & 0xf;
instn = *pc++;
instn_category = (instn >> 12) & 0xf;
data1 = instn & 0xff;
type2 = (instn >> 8) & 0xf;
fprintf(f_out, " %.3x: %.4x %s", (unsigned int)(pc-prog_buf), instn, instn_category_string[instn_category]);

switch (instn_category)
@@ -503,7 +510,6 @@ static void DecompilePhoneme(FILE *f_out, PHONEME_TAB *ph, int compile_phoneme)
data1 = 0;
fprintf(f_out, "%s", instn0_string[data1]);
} else if (type2 == i_IPA_NAME) {
int ix;
for (ix = 0; ix < data1; ix += 2) {
instn = *pc++;
buf[ix] = instn >> 8;
@@ -565,8 +571,8 @@ static void DecompilePhoneme(FILE *f_out, PHONEME_TAB *ph, int compile_phoneme)
} else if (type2 == 8) {
// list of numbers
fprintf(f_out, " StressLevel(");
int any = 0;
for (int ix = 0; ix < 8; ix++) {
any = 0;
for (ix = 0; ix < 8; ix++) {
if (data1 & (1 << ix)) {
if (any)
fputc(',', f_out);
@@ -703,10 +709,12 @@ char item_string[N_ITEM_STRING];

static int ref_sorter(char **a, char **b)
{
int ix;

REF_HASH_TAB *p1 = (REF_HASH_TAB *)(*a);
REF_HASH_TAB *p2 = (REF_HASH_TAB *)(*b);

int ix = strcoll(p1->string, p2->string);
ix = strcoll(p1->string, p2->string);
if (ix != 0)
return ix;

@@ -720,9 +728,14 @@ static int ref_sorter(char **a, char **b)
static void CompileReport(void)
{
int ix;
int hash;
int n;
REF_HASH_TAB *p;
REF_HASH_TAB **list;
const char *data_path;
int prev_table;
int procedure_num;
int prev_mnemonic;

if (f_report == NULL)
return;
@@ -740,19 +753,19 @@ static void CompileReport(void)

fprintf(f_report, "Data file Used by\n");
ix = 0;
for (int hash = 0; (hash < 256) && (ix < count_references); hash++) {
for (hash = 0; (hash < 256) && (ix < count_references); hash++) {
p = ref_hash_tab[hash];
while (p != NULL) {
list[ix++] = p;
p = (REF_HASH_TAB *)(p->link);
}
}
int n = ix;
n = ix;
qsort((void *)list, n, sizeof(REF_HASH_TAB *), (int (*)(const void *, const void *))ref_sorter);

data_path = "";
int prev_mnemonic = 0;
int prev_table = 0;
prev_mnemonic = 0;
prev_table = 0;
for (ix = 0; ix < n; ix++) {
int j = 0;

@@ -771,7 +784,7 @@ static void CompileReport(void)
prev_mnemonic = list[ix]->ph_mnemonic;
if ((prev_mnemonic >> 24) == 'P') {
// a procedure, not a phoneme
int procedure_num = atoi(WordToString(prev_mnemonic));
procedure_num = atoi(WordToString(prev_mnemonic));
fprintf(f_report, " %s %s", phoneme_tab_list2[prev_table = list[ix]->ph_table].name, proc_names[procedure_num]);
} else
fprintf(f_report, " [%s] %s", WordToString(prev_mnemonic), phoneme_tab_list2[prev_table = list[ix]->ph_table].name);
@@ -818,14 +831,15 @@ static FILE *fopen_log(FILE *f_log, const char *fname, const char *access)
static unsigned int StringToWord(const char *string)
{
// Pack 4 characters into a word
int ix;
unsigned char c;
unsigned int word;

if (string == NULL)
return 0;

unsigned int word = 0;
for (int ix = 0; ix < 4; ix++) {
word = 0;
for (ix = 0; ix < 4; ix++) {
if (string[ix] == 0) break;
c = string[ix];
word |= (c << (ix*8));
@@ -895,16 +909,21 @@ static int LookupPhoneme(const char *string, int control)
// control = 1 declare phoneme if not found
// control = 2 start looking after control & stress phonemes

int ix;
int start;
int use;
unsigned int word;

if (strcmp(string, "NULL") == 0)
return 1;

int ix = strlen(string);
ix = strlen(string);
if ((ix == 0) || (ix > 4))
error("Bad phoneme name '%s'", string);
unsigned int word = StringToWord(string);
word = StringToWord(string);

// don't use phoneme number 0, reserved for string terminator
int start = 1;
start = 1;

if (control == 2) {
// don't look for control and stress phonemes (allows these characters to be
@@ -912,7 +931,7 @@ static int LookupPhoneme(const char *string, int control)
start = 8;
}

int use = 0;
use = 0;
for (ix = start; ix < n_phcodes; ix++) {
if (phoneme_tab2[ix].mnemonic == word)
return ix;
@@ -938,7 +957,8 @@ static int LookupPhoneme(const char *string, int control)

static unsigned int get_char()
{
unsigned int c = fgetc(f_in);
unsigned int c;
c = fgetc(f_in);
if (c == '\n')
linenum++;
return c;
@@ -965,6 +985,8 @@ static int NextItem(int type)
int acc;
unsigned char c = 0;
unsigned char c2;
int ix;
int sign;
char *p;
keywtab_t *pk;

@@ -995,7 +1017,7 @@ static int NextItem(int type)
return -1;
}

int ix = 0;
ix = 0;
while (!feof(f_in) && !isspace(c) && (c != '(') && (c != ')') && (c != ',')) {
if (c == '\\')
c = get_char();
@@ -1026,7 +1048,7 @@ static int NextItem(int type)

if ((type == tNUMBER) || (type == tSIGNEDNUMBER)) {
acc = 0;
int sign = 1;
sign = 1;
p = item_string;

if ((*p == '-') && (type == tSIGNEDNUMBER)) {
@@ -1067,9 +1089,10 @@ static int NextItem(int type)
static int NextItemMax(int max)
{
// Get a number, but restrict value to max
int value;
char msg[80];

int value = NextItem(tNUMBER);
value = NextItem(tNUMBER);
if (value > max) {
sprintf(msg, "Value %d is greater than maximum %d", value, max);
error(msg, NULL);
@@ -1084,12 +1107,14 @@ static int NextItemBrackets(int type, int control)
// control: bit 0 0= need (
// bit 1 1= allow comma

int value;

if ((control & 1) == 0) {
if (!NextItem(tOPENBRACKET))
error("Expected '('", NULL);
}

int value = NextItem(type);
value = NextItem(type);
if ((control & 2) && (item_terminator == ','))
return value;

@@ -1122,7 +1147,7 @@ static int Range(int value, int divide, int min, int max)
int CompileVowelTransition(int which)
{
// Compile a vowel transition
int key;
int len = 0;
int rms = 0;
int f1 = 0;
@@ -1133,7 +1158,10 @@ int CompileVowelTransition(int which)
int f3_amp = 0;
int flags = 0;
int vcolour = 0;
int x;
int instn = i_VOWELIN;
int word1;
int word2;

if (which == 1) {
len = 50 / 2; // defaults for transition into vowel
@@ -1150,7 +1178,7 @@ int CompileVowelTransition(int which)
}

for (;;) {
int key = NextItem(tKEYWORD);
key = NextItem(tKEYWORD);
if (item_type != tTRANSITION) {
UngetItem();
break;
@@ -1174,7 +1202,7 @@ int CompileVowelTransition(int which)
f2_min = Range(NextItem(tSIGNEDNUMBER), 50, -15, 15) & 0x1f;
f2_max = Range(NextItem(tSIGNEDNUMBER), 50, -15, 15) & 0x1f;
if (f2_min > f2_max) {
int x = f2_min;
x = f2_min;
f2_min = f2_max;
f2_max = x;
}
@@ -1211,8 +1239,8 @@ int CompileVowelTransition(int which)
break;
}
}
int word1 = len + (rms << 6) + (flags << 12);
int word2 = f2 + (f2_min << 6) + (f2_max << 11) + (f3_adj << 16) + (f3_amp << 21) + (f1 << 26) + (vcolour << 29);
word1 = len + (rms << 6) + (flags << 12);
word2 = f2 + (f2_min << 6) + (f2_max << 11) + (f3_adj << 16) + (f3_amp << 21) + (f1 << 26) + (vcolour << 29);
prog_out[0] = instn + ((word1 >> 16) & 0xff);
prog_out[1] = word1;
prog_out[2] = word2 >> 16;
@@ -1225,8 +1253,14 @@ int CompileVowelTransition(int which)
int LoadSpect(const char *path, int control)
{
SpectSeq *spectseq;
int peak;
int displ;
int frame;
int n_frames;
int ix;
int x, x2;
int rms;
float total;
float pkheight;
int marker1_set = 0;
int frame_vowelbreak = 0;
@@ -1255,21 +1289,21 @@ int LoadSpect(const char *path, int control)
}

// do we need additional klatt data ?
for (int frame = 0; frame < spectseq->numframes; frame++) {
for (frame = 0; frame < spectseq->numframes; frame++) {
for (ix = 5; ix < N_KLATTP2; ix++) {
if (spectseq->frames[frame]->klatt_param[ix] != 0)
klatt_flag = FRFLAG_KLATT;
}
}

int displ = ftell(f_phdata);
displ = ftell(f_phdata);

seq_out.n_frames = 0;
seq_out.sqflags = 0;
seq_out.length_total = 0;

float total = 0.0f;
for (int frame = 0; frame < spectseq->numframes; frame++) {
total = 0;
for (frame = 0; frame < spectseq->numframes; frame++) {
if (spectseq->frames[frame]->keyframe) {
if (seq_out.n_frames == 1)
frame_vowelbreak = frame;
@@ -1291,8 +1325,8 @@ int LoadSpect(const char *path, int control)
spectseq->frames[frame_vowelbreak]->markers |= FRFLAG_VOWEL_CENTRE;
}

int n_frames = 0;
for (int frame = 0; frame < spectseq->numframes; frame++) {
n_frames = 0;
for (frame = 0; frame < spectseq->numframes; frame++) {
fr = spectseq->frames[frame];

if (fr->keyframe) {
@@ -1307,7 +1341,7 @@ int LoadSpect(const char *path, int control)

fr_out->frflags = fr->markers | klatt_flag;

int rms = (int)GetFrameRms(fr, spectseq->amplitude);
rms = (int)GetFrameRms(fr, spectseq->amplitude);
if (rms > 255) rms = 255;
fr_out->rms = rms;

@@ -1316,7 +1350,7 @@ int LoadSpect(const char *path, int control)

// write: peak data
count_frames++;
for (int peak = 0; peak < 8; peak++) {
for (peak = 0; peak < 8; peak++) {
if (peak < 7)
fr_out->ffreq[peak] = fr->peaks[peak].pkfreq;

@@ -1355,7 +1389,7 @@ int LoadSpect(const char *path, int control)
for (ix = 0; ix < 5; ix++)
fr_out->klattp2[ix] = fr->klatt_param[ix+5];

for (int peak = 0; peak < 7; peak++) {
for (peak = 0; peak < 7; peak++) {
fr_out->klatt_ap[ix] = fr->peaks[peak].klt_ap;

x = fr->peaks[peak].klt_bp / 2;
@@ -1395,13 +1429,18 @@ int LoadSpect(const char *path, int control)

static int LoadWavefile(FILE *f, const char *fname)
{
int displ;
unsigned char c1;
unsigned char c3;
int c2;
int sample;
int sample2;
float x;
int max = 0;
int length;
int sr1, sr2;
int failed;
int len;
int resample_wav = 0;
const char *fname2;
char fname_temp[100];
@@ -1409,8 +1448,8 @@ static int LoadWavefile(FILE *f, const char *fname)
int scale_factor = 0;

fseek(f, 24, SEEK_SET);
int sr1 = Read4Bytes(f);
int sr2 = Read4Bytes(f);
sr1 = Read4Bytes(f);
sr2 = Read4Bytes(f);
fseek(f, 40, SEEK_SET);

if ((sr1 != samplerate_native) || (sr2 != sr1*2)) {
@@ -1428,7 +1467,7 @@ static int LoadWavefile(FILE *f, const char *fname)
#endif

fname2 = fname;
size_t len = strlen(fname);
len = strlen(fname);
if (strcmp(&fname[len-4], ".wav") == 0) {
strcpy(msg, fname);
msg[len-4] = 0;
@@ -1465,10 +1504,10 @@ static int LoadWavefile(FILE *f, const char *fname)
fseek(f, 40, SEEK_SET); // skip past the WAV header, up to before "data length"
}

int displ = ftell(f_phdata);
displ = ftell(f_phdata);

// data contains: 4 bytes of length (n_samples * 2), followed by 2-byte samples (lsb byte first)
int length = Read4Bytes(f);
length = Read4Bytes(f);

while (!feof(f)) {
c1 = fgetc(f);
@@ -1510,7 +1549,7 @@ static int LoadWavefile(FILE *f, const char *fname)
fputc(sample & 0xff, f_phdata);
fputc(sample >> 8, f_phdata);
} else {
float x = ((float)sample / scale_factor) + 0.5;
x = ((float)sample / scale_factor) + 0.5;
sample2 = (int)x;
if (sample2 > 127)
sample2 = 127;
@@ -1536,9 +1575,10 @@ static int LoadWavefile(FILE *f, const char *fname)

static int LoadEnvelope(FILE *f, const char *fname)
{
int displ;
char buf[128];

int displ = ftell(f_phdata);
displ = ftell(f_phdata);

fseek(f, 12, SEEK_SET);
if (fread(buf, 128, 1, f) == 0)
@@ -1573,16 +1613,20 @@ static int Hash8(const char *string)

static int LoadEnvelope2(FILE *f, const char *fname)
{
int ix, ix2;
int n;
int x, y;
int displ;
int n_points;
double yy;
char line_buf[128];
float env_x[20];
float env_y[20];
int env_lin[20];
unsigned char env[ENV_LEN];

int n_points = 0;
n_points = 0;
fgets(line_buf, sizeof(line_buf), f); // skip first line

while (!feof(f)) {
if (fgets(line_buf, sizeof(line_buf), f) == NULL)
break;
@@ -1597,17 +1641,16 @@ static int LoadEnvelope2(FILE *f, const char *fname)
env_x[n_points] = env_x[n_points-1];
env_y[n_points] = env_y[n_points-1];

int ix = -1;
int ix2 = 0;
int y = 0;
for (int x = 0; x < ENV_LEN; x++) {
ix = -1;
ix2 = 0;
for (x = 0; x < ENV_LEN; x++) {
if (x > env_x[ix+4])
ix++;
if (x >= env_x[ix2+1])
ix2++;

if (env_lin[ix2] > 0) {
double yy = env_y[ix2] + (env_y[ix2+1] - env_y[ix2]) * ((float)x - env_x[ix2]) / (env_x[ix2+1] - env_x[ix2]);
yy = env_y[ix2] + (env_y[ix2+1] - env_y[ix2]) * ((float)x - env_x[ix2]) / (env_x[ix2+1] - env_x[ix2]);
y = (int)(yy * 2.55);
} else if (n_points > 3)
y = (int)(polint(&env_x[ix], &env_y[ix], 4, x) * 2.55); // convert to range 0-255
@@ -1624,7 +1667,7 @@ static int LoadEnvelope2(FILE *f, const char *fname)
n_envelopes++;
}

int displ = ftell(f_phdata);
displ = ftell(f_phdata);
fwrite(env, 1, 128, f_phdata);

return displ;
@@ -1636,6 +1679,8 @@ static int LoadDataFile(const char *path, int control)
// return index into spect or sample data area. bit 23=1 if a sample

FILE *f;
int id;
int hash;
int addr = 0;
int type_code = ' ';
REF_HASH_TAB *p, *p2;
@@ -1648,7 +1693,7 @@ static int LoadDataFile(const char *path, int control)

count_references++;

int hash = Hash8(path);
hash = Hash8(path);
p = ref_hash_tab[hash];
while (p != NULL) {
if (strcmp(path, p->string) == 0) {
@@ -1670,7 +1715,7 @@ static int LoadDataFile(const char *path, int control)
}
}

int id = Read4Bytes(f);
id = Read4Bytes(f);
rewind(f);

if (id == 0x43455053) {
@@ -1712,11 +1757,13 @@ static int LoadDataFile(const char *path, int control)

static int CompileToneSpec(void)
{
int pitch1 = 0;
int pitch2 = 0;
int pitch_env = 0;
int amp_env = 0;

int pitch1 = NextItemBrackets(tNUMBER, 2);
int pitch2 = NextItemBrackets(tNUMBER, 3);
pitch1 = NextItemBrackets(tNUMBER, 2);
pitch2 = NextItemBrackets(tNUMBER, 3);

if (item_terminator == ',') {
NextItemBrackets(tSTRING, 3);
@@ -1752,6 +1799,7 @@ static int CompileToneSpec(void)

int CompileSound(int keyword, int isvowel)
{
int addr;
int value = 0;
char path[N_ITEM_STRING];
static int sound_instns[] = { i_FMT, i_WAV, i_VWLSTART, i_VWLENDING, i_WAVADD };
@@ -1777,7 +1825,7 @@ int CompileSound(int keyword, int isvowel)
}
}
}
int addr = LoadDataFile(path, isvowel);
addr = LoadDataFile(path, isvowel);
addr = addr / 4; // addr is words not bytes

*prog_out++ = sound_instns[keyword-kFMT] + ((value & 0xff) << 4) + ((addr >> 16) & 0xf);
@@ -1803,6 +1851,8 @@ int CompileIf(int elif)
int word = 0;
int word2;
int data;
int bitmap;
int brackets;
int not_flag;
USHORT *prog_last_if = NULL;

@@ -1844,8 +1894,8 @@ int CompileIf(int elif)
}
} else if (item_type == tTEST) {
if (key == kTHISSTRESS) {
int bitmap = 0;
int brackets = 2;
bitmap = 0;
brackets = 2;
do {
data = NextItemBrackets(tNUMBER, brackets);
if (data > 7)
@@ -1906,10 +1956,11 @@ int CompileIf(int elif)
void FillThen(int add)
{
USHORT *p;
int offset;

p = if_stack[if_level].p_then;
if (p != NULL) {
int offset = prog_out - p + add;
offset = prog_out - p + add;

if ((then_count == 1) && (if_level == 1)) {
// The THEN part only contains one statement, we can remove the THEN jump
@@ -2019,7 +2070,9 @@ static int CompileSwitch(int type)

static PHONEME_TAB_LIST *FindPhonemeTable(const char *string)
{
for (int ix = 0; ix < n_phoneme_tabs; ix++) {
int ix;

for (ix = 0; ix < n_phoneme_tabs; ix++) {
if (strcmp(phoneme_tab_list2[ix].name, string) == 0)
return &phoneme_tab_list2[ix];
}
@@ -2031,6 +2084,7 @@ static PHONEME_TAB *FindPhoneme(const char *string)
{
PHONEME_TAB_LIST *phtab = NULL;
int ix;
unsigned int mnem;
char *phname;
char buf[200];

@@ -2047,7 +2101,7 @@ static PHONEME_TAB *FindPhoneme(const char *string)
if (phtab == NULL)
return NULL; // phoneme table not found

unsigned int mnem = StringToWord(phname);
mnem = StringToWord(phname);
for (ix = 1; ix < 256; ix++) {
if (mnem == phtab->phoneme_tab_ptr[ix].mnemonic)
return &phtab->phoneme_tab_ptr[ix];
@@ -2059,6 +2113,8 @@ static PHONEME_TAB *FindPhoneme(const char *string)

static void ImportPhoneme(void)
{
unsigned int ph_mnem;
unsigned int ph_code;
PHONEME_TAB *ph;

NextItem(tSTRING);
@@ -2066,8 +2122,8 @@ static void ImportPhoneme(void)
if ((ph = FindPhoneme(item_string)) == NULL)
return;

unsigned int ph_mnem = phoneme_out->mnemonic;
unsigned int ph_code = phoneme_out->code;
ph_mnem = phoneme_out->mnemonic;
ph_code = phoneme_out->code;
memcpy(phoneme_out, ph, sizeof(PHONEME_TAB));
phoneme_out->mnemonic = ph_mnem;
phoneme_out->code = ph_code;
@@ -2120,7 +2176,9 @@ static void DecThenCount()

static void InstnPlusPhoneme(int instn)
{
int phcode = NextItemBrackets(tPHONEMEMNEM, 0);
int phcode;

phcode = NextItemBrackets(tPHONEMEMNEM, 0);
*prog_out++ = instn + phcode;
}

@@ -2130,6 +2188,10 @@ int CompilePhoneme(int compile_phoneme)
int keyword;
int value;
int phcode = 0;
int flags;
int ix;
int start;
int count;
int c;
char *p;
int vowel_length_factor = 100; // for testing
@@ -2251,9 +2313,9 @@ int CompilePhoneme(int compile_phoneme)
strcpy(item_string, " ");

// copy the string, recognize characters in the form U+9999
int flags = 0;
int count = 0;
int ix = 1;
flags = 0;
count = 0;
ix = 1;

for (p = item_string; *p != 0;) {
p += utf8_in(&c, p);
@@ -2286,7 +2348,7 @@ int CompilePhoneme(int compile_phoneme)
ipa_buf[0] = flags;
ipa_buf[ix] = 0;

int start = 1;
start = 1;
if (flags != 0)
start = 0; // only include the flags byte if bits are set
value = strlen(&ipa_buf[start]); // number of UTF-8 bytes
@@ -2465,24 +2527,27 @@ int CompilePhoneme(int compile_phoneme)

static void WritePhonemeTables()
{
int ix;
int j;
int n;
int value;
int count;
PHONEME_TAB *p;

int value = n_phoneme_tabs;
value = n_phoneme_tabs;
fputc(value, f_phtab);
fputc(0, f_phtab);
fputc(0, f_phtab);
fputc(0, f_phtab);

for (int ix = 0; ix < n_phoneme_tabs; ix++) {
for (ix = 0; ix < n_phoneme_tabs; ix++) {
p = phoneme_tab_list2[ix].phoneme_tab_ptr;
n = n_phcodes_list[ix];
p[n].mnemonic = 0; // terminate the phoneme table

// count number of locally declared phonemes
count = 0;
for (int j = 0; j < n; j++) {
for (j = 0; j < n; j++) {
if (ix == 0)
p[j].phflags |= phLOCAL; // write all phonemes in the base phoneme table

@@ -2499,7 +2564,7 @@ static void WritePhonemeTables()

fwrite(phoneme_tab_list2[ix].name, 1, N_PHONEME_TAB_NAME, f_phtab);

for (int j = 0; j < n; j++) {
for (j = 0; j < n; j++) {
if (p[j].phflags & phLOCAL) {
// this bit is set temporarily to incidate a local phoneme, declared in
// in the current phoneme file
@@ -2551,6 +2616,7 @@ static void EndPhonemeTable()
static void StartPhonemeTable(const char *name)
{
int ix;
int j;
PHONEME_TAB *p;

fprintf(f_errors, "______________________________\nPhoneme Table: '%s'\n", name);
@@ -2584,7 +2650,7 @@ static void StartPhonemeTable(const char *name)
n_phcodes = n_phcodes_list[ix];

// clear "local phoneme" bit"
for (int j = 0; j < n_phcodes; j++)
for (j = 0; j < n_phcodes; j++)
phoneme_tab2[j].phflags &= ~phLOCAL;
break;
}
@@ -2601,6 +2667,7 @@ static void CompileEquivalents()
{
// a list of phonemes in another language and the equivalent phoneme strings in this language

int ix;
int n_names;
int n_bytes;
int foreign_phoneme;
@@ -2638,7 +2705,7 @@ static void CompileEquivalents()
if ((p = strstr(line_buf, "//")) != NULL)
*p = 0;

for (int ix = 0; ix < 6; ix++)
for (ix = 0; ix < 6; ix++)
names[ix][0] = 0;
n_names = sscanf(line_buf, "%s %s %s %s %s %s", names[0], names[1], names[2], names[3], names[4], names[5]);
if (n_names < 1)
@@ -2668,7 +2735,7 @@ static void CompileEquivalents()
}
}

for (int ix = 1; ix < n_names; ix++)
for (ix = 1; ix < n_names; ix++)
phcode[ix] = LookupPhoneme(names[ix], 1);

// only write a translation if it has an effect

+ 52
- 32
src/libespeak-ng/compiledict.c View File

@@ -197,6 +197,7 @@ const char *LookupMnemName(MNEM_TAB *table, const int value)
void print_dictionary_flags(unsigned int *flags, char *buf, int buf_len)
{
int stress;
int ix;
const char *name;
int len;
int total = 0;
@@ -208,7 +209,7 @@ void print_dictionary_flags(unsigned int *flags, char *buf, int buf_len)
buf += total;
}

for (int ix = 8; ix < 64; ix++) {
for (ix = 8; ix < 64; ix++) {
if (((ix < 30) && (flags[0] & (1 << ix))) || ((ix >= 0x20) && (flags[1] & (1 << (ix-0x20))))) {
name = LookupMnemName(mnem_flags, ix);
len = strlen(name) + 1;
@@ -229,6 +230,8 @@ char *DecodeRule(const char *group_chars, int group_length, char *rule, int cont
unsigned char c;
char *p;
char *p_end;
int ix;
int match_type;
int finished = 0;
int value;
int linenum = 0;
@@ -251,10 +254,9 @@ char *DecodeRule(const char *group_chars, int group_length, char *rule, int cont

static char symbols_lg[] = { 'A', 'B', 'C', 'H', 'F', 'G', 'Y' };

int match_type = 0;
match_type = 0;
buf_pre[0] = 0;

int ix;
for (ix = 0; ix < group_length; ix++)
buf[ix] = group_chars[ix];
buf[ix] = 0;
@@ -386,26 +388,35 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
char *word;
char *phonetic;
unsigned int ix;
int step;
unsigned int n_flag_codes = 0;
int flagnum;
int flag_offset;
int length;
int multiple_words = 0;
int multiple_numeric_hyphen = 0;
char *multiple_string = NULL;
char *multiple_string_end = NULL;

int len_word;
int len_phonetic;
int text_not_phonemes; // this word specifies replacement text, not phonemes
unsigned int wc;
int all_upper_case;

char *mnemptr;
unsigned char flag_codes[100];
char encoded_ph[200];
char bad_phoneme_str[4];
int bad_phoneme;
static char nullstring[] = { 0 };

int text_not_phonemes = 0; // this word specifies replacement text, not phonemes
text_not_phonemes = 0;
phonetic = word = nullstring;

p = linebuf;

int step = 0;
step = 0;

c = 0;
while (c != '\n') {
@@ -413,7 +424,7 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)

if ((c == '?') && (step == 0)) {
// conditional rule, allow only if the numbered condition is set for the voice
int flag_offset = 100;
flag_offset = 100;

p++;
if (*p == '!') {
@@ -441,7 +452,7 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
while (!isspace2(c = *p)) p++;
*p = 0;

int flagnum = LookupMnem(mnem_flags, mnemptr);
flagnum = LookupMnem(mnem_flags, mnemptr);
if (flagnum > 0) {
if (flagnum == 200)
text_mode = 1;
@@ -552,14 +563,12 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
// this is replacement text, so don't encode as phonemes. Restrict the length of the replacement word
strncpy0(encoded_ph, phonetic, N_WORD_BYTES-4);
} else {
int bad_phoneme;
EncodePhonemes(phonetic, encoded_ph, &bad_phoneme);
if (strchr(encoded_ph, phonSWITCH) != 0)
flag_codes[n_flag_codes++] = BITNUM_FLAG_ONLY_S; // don't match on suffixes (except 's') when switching languages

// check for errors in the phonemes codes
if (bad_phoneme != 0) {
char bad_phoneme_str[4];
// unrecognised phoneme, report error
bad_phoneme_str[utf8_out(bad_phoneme, bad_phoneme_str)] = 0;
fprintf(f_log, "%5d: Bad phoneme [%s] (U+%x) in: %s %s\n", linenum, bad_phoneme_str, bad_phoneme, word, phonetic);
@@ -578,7 +587,7 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
// convert to lower case, and note if the word is all-capitals
int c2;

int all_upper_case = 1;
all_upper_case = 1;
p = word;
for (p = word;;) {
// this assumes that the lower case char is the same length as the upper case char
@@ -596,13 +605,13 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
flag_codes[n_flag_codes++] = BITNUM_FLAG_ALLCAPS;
}

int len_word = strlen(word);
len_word = strlen(word);

if (translator->transpose_min > 0)
len_word = TransposeAlphabet(translator, word);

*hash = HashDictionary(word);
int len_phonetic = strlen(encoded_ph);
len_phonetic = strlen(encoded_ph);

dict_line[1] = len_word; // bit 6 indicates whether the word has been compressed
len_word &= 0x3f;
@@ -643,10 +652,11 @@ static int compile_line(char *linebuf, char *dict_line, int *hash)
static void compile_dictlist_start(void)
{
// initialise dictionary list
int ix;
char *p;
char *p2;

for (int ix = 0; ix < N_HASH_DICT; ix++) {
for (ix = 0; ix < N_HASH_DICT; ix++) {
p = hash_chains[ix];
while (p != NULL) {
memcpy(&p2, p, sizeof(char *));
@@ -661,10 +671,11 @@ static void compile_dictlist_start(void)
static void compile_dictlist_end(FILE *f_out)
{
// Write out the compiled dictionary list
int hash;
int length;
char *p;

for (int hash = 0; hash < N_HASH_DICT; hash++) {
for (hash = 0; hash < N_HASH_DICT; hash++) {
p = hash_chains[hash];
hash_counts[hash] = (int)ftell(f_out);

@@ -684,6 +695,7 @@ static int compile_dictlist_file(const char *path, const char *filename)
char *p;
int count = 0;
FILE *f_in;
char buf[200];
char fname[sizeof(path_home)+45];
char dict_line[128];

@@ -702,7 +714,6 @@ static int compile_dictlist_file(const char *path, const char *filename)

linenum = 0;

char buf[200];
while (fgets(buf, sizeof(buf), f_in) != NULL) {
linenum++;

@@ -764,6 +775,7 @@ static void copy_rule_string(char *string, int *state_out)
int len;
char c;
int c2, c3;
int sxflags;
int value;
int literal;
int hexdigit_input = 0;
@@ -780,8 +792,7 @@ static void copy_rule_string(char *string, int *state_out)
rule_phonemes[len++] = ' ';
output = &rule_phonemes[len];
}

int sxflags = 0x808000; // to ensure non-zero bytes
sxflags = 0x808000; // to ensure non-zero bytes

for (p = string, ix = 0;;) {
literal = 0;
@@ -997,12 +1008,18 @@ static char *compile_rule(char *input)
{
int ix;
unsigned char c;
int wc;
char *p;
char *prule;
int len;
int len_name;
int start;
int state = 2;
int finish = 0;
char buf[80];
char output[150];
int bad_phoneme;
char bad_phoneme_str[4];

buf[0] = 0;
rule_cond[0] = 0;
@@ -1071,8 +1088,6 @@ static char *compile_rule(char *input)
return NULL;
}

int bad_phoneme;
char bad_phoneme_str[4];
EncodePhonemes(rule_phonemes, buf, &bad_phoneme);
if (bad_phoneme != 0) {
bad_phoneme_str[utf8_out(bad_phoneme, bad_phoneme_str)] = 0;
@@ -1080,11 +1095,10 @@ static char *compile_rule(char *input)
error_count++;
}
strcpy(output, buf);
size_t len = strlen(buf)+1;
len = strlen(buf)+1;

size_t len_name = strlen(group_name);
len_name = strlen(group_name);
if ((len_name > 0) && (memcmp(rule_match, group_name, len_name) != 0)) {
int wc;
utf8_in(&wc, rule_match);
if ((group_name[0] == '9') && IsDigit(wc)) {
// numeric group, rule_match starts with a digit, so OK
@@ -1122,7 +1136,7 @@ static char *compile_rule(char *input)
}
}
if (rule_pre[0] != 0) {
int start = 0;
start = 0;
if (rule_pre[0] == RULE_SPACE) {
// omit '_' at the beginning of the pre-string and imply it by using RULE_PRE_ATSTART
c = RULE_PRE_ATSTART;
@@ -1161,8 +1175,8 @@ int __cdecl string_sorter(char **a, char **b)
static int __cdecl rgroup_sorter(RGROUP *a, RGROUP *b)
{
// Sort long names before short names
int ix = strlen(b->name) - strlen(a->name);
int ix;
ix = strlen(b->name) - strlen(a->name);
if (ix != 0) return ix;
ix = strcmp(a->name, b->name);
if (ix != 0) return ix;
@@ -1171,8 +1185,10 @@ static int __cdecl rgroup_sorter(RGROUP *a, RGROUP *b)

static void output_rule_group(FILE *f_out, int n_rules, char **rules, char *name)
{
int ix;
int len1;
int len2;
int len_name;
char *p;
char *p2, *p3;
const char *common;
@@ -1180,7 +1196,7 @@ static void output_rule_group(FILE *f_out, int n_rules, char **rules, char *name
short nextchar_count[256];
memset(nextchar_count, 0, sizeof(nextchar_count));

int len_name = strlen(name);
len_name = strlen(name);

// sort the rules in this group by their phoneme string
common = "";
@@ -1189,7 +1205,7 @@ static void output_rule_group(FILE *f_out, int n_rules, char **rules, char *name
if (strcmp(name, "9") == 0)
len_name = 0; // don't remove characters from numeric match strings

for (int ix = 0; ix < n_rules; ix++) {
for (ix = 0; ix < n_rules; ix++) {
p = rules[ix];
len1 = strlen(p) + 1; // phoneme string
p3 = &p[len1];
@@ -1218,6 +1234,9 @@ static int compile_lettergroup(char *input, FILE *f_out)
{
char *p;
char *p_start;
int group;
int ix;
int n_items;
int length;
int max_length = 0;

@@ -1232,7 +1251,7 @@ static int compile_lettergroup(char *input, FILE *f_out)
return 1;
}

int group = atoi(&p[0]);
group = atoi(&p[0]);
if (group >= N_LETTER_GROUPS) {
fprintf(f_log, "%5d: lettergroup out of range (01-%.2d)\n", linenum, N_LETTER_GROUPS-1);
error_count++;
@@ -1250,7 +1269,7 @@ static int compile_lettergroup(char *input, FILE *f_out)
}
letterGroupsDefined[group] = 1;

int n_items = 0;
n_items = 0;
while (n_items < N_LETTERGP_ITEMS) {
while (isspace2(*p)) p++;
if (*p == 0)
@@ -1270,7 +1289,7 @@ static int compile_lettergroup(char *input, FILE *f_out)

// write out the items, longest first
while (max_length > 1) {
for (int ix = 0; ix < n_items; ix++) {
for (ix = 0; ix < n_items; ix++) {
if (item_length[ix] == max_length)
fwrite(items[ix], 1, max_length, f_out);
}
@@ -1498,6 +1517,8 @@ int CompileDictionary(const char *dsource, const char *dict_name, FILE *log, cha

FILE *f_in;
FILE *f_out;
int offset_rules = 0;
int value;
char fname_in[sizeof(path_home)+45];
char fname_out[sizeof(path_home)+15];
char fname_temp[sizeof(path_home)+15];
@@ -1537,8 +1558,7 @@ int CompileDictionary(const char *dsource, const char *dict_name, FILE *log, cha
}
sprintf(fname_temp, "%s%ctemp", path_home, PATHSEP);

int value = N_HASH_DICT;
int offset_rules = 0;
value = N_HASH_DICT;
Write4Bytes(f_out, value);
Write4Bytes(f_out, offset_rules);


+ 2
- 1
src/libespeak-ng/compilembrola.c View File

@@ -44,6 +44,7 @@ static const char *basename(const char *filename)
static unsigned int StringToWord(const char *string)
{
// Pack 4 characters into a word
int ix;
unsigned char c;
unsigned int word;

@@ -51,7 +52,7 @@ static unsigned int StringToWord(const char *string)
return 0;

word = 0;
for (int ix = 0; ix < 4; ix++) {
for (ix = 0; ix < 4; ix++) {
if (string[ix] == 0) break;
c = string[ix];
word |= (c << (ix*8));

+ 57
- 29
src/libespeak-ng/dictionary.c View File

@@ -92,9 +92,10 @@ int Reverse4Bytes(int word)
{
// reverse the order of bytes from little-endian to big-endian
#ifdef ARCH_BIG
int ix;
int word2 = 0;

for (int ix = 0; ix <= 24; ix += 8) {
for (ix = 0; ix <= 24; ix += 8) {
word2 = word2 << 8;
word2 |= (word >> ix) & 0xff;
}
@@ -200,9 +201,12 @@ static void InitGroups(Translator *tr)

int LoadDictionary(Translator *tr, const char *name, int no_error)
{
int hash;
char *p;
int *pw;
int length;
FILE *f;
unsigned int size;
char fname[sizeof(path_home)+20];

strcpy(dictionary_name, name); // currently loaded dictionary name
@@ -212,7 +216,7 @@ int LoadDictionary(Translator *tr, const char *name, int no_error)
// bytes 0-3: offset to rules data
// bytes 4-7: number of hash table entries
sprintf(fname, "%s%c%s_dict", path_home, PATHSEP, name);
unsigned int size = GetFileLength(fname);
size = GetFileLength(fname);

if (tr->data_dictlist != NULL) {
Free(tr->data_dictlist);
@@ -233,7 +237,7 @@ int LoadDictionary(Translator *tr, const char *name, int no_error)
fclose(f);

pw = (int *)(tr->data_dictlist);
int length = Reverse4Bytes(pw[1]);
length = Reverse4Bytes(pw[1]);

if (size <= (N_HASH_DICT + sizeof(int)*2)) {
fprintf(stderr, "Empty _dict file: '%s\n", fname);
@@ -253,7 +257,7 @@ int LoadDictionary(Translator *tr, const char *name, int no_error)
// set up hash table for data_dictlist
p = &(tr->data_dictlist[8]);

for (int hash = 0; hash < N_HASH_DICT; hash++) {
for (hash = 0; hash < N_HASH_DICT; hash++) {
tr->dict_hashtab[hash] = p;
while ((length = *p) != 0)
p += length;
@@ -292,6 +296,7 @@ int HashDictionary(const char *string)
*/
const char *EncodePhonemes(const char *p, char *outptr, int *bad_phoneme)
{
int ix;
unsigned char c;
int count; // num. of matching characters
int max; // highest num. of matching found so far
@@ -328,7 +333,7 @@ const char *EncodePhonemes(const char *p, char *outptr, int *bad_phoneme)
max = -1;
max_ph = 0;

for (int ix = 1; ix < n_phoneme_tab; ix++) {
for (ix = 1; ix < n_phoneme_tab; ix++) {
if (phoneme_tab[ix] == NULL)
continue;
if (phoneme_tab[ix]->type == phINVALID)
@@ -437,6 +442,9 @@ static unsigned int phon_out_size = 0;
char *WritePhMnemonic(char *phon_out, PHONEME_TAB *ph, PHONEME_LIST *plist, int use_ipa, int *flags)
{
int c;
int mnem;
int len;
int first;
int ix = 0;
char *p;
PHONEME_DATA phdata;
@@ -476,7 +484,7 @@ char *WritePhMnemonic(char *phon_out, PHONEME_TAB *ph, PHONEME_LIST *plist, int
p++;
}

int len = strlen(p);
len = strlen(p);
if (len > 0) {
strcpy(phon_out, p);
phon_out += len;
@@ -485,8 +493,8 @@ char *WritePhMnemonic(char *phon_out, PHONEME_TAB *ph, PHONEME_LIST *plist, int
}
}

int first = 1;
for (int mnem = ph->mnemonic; (c = mnem & 0xff) != 0; mnem = mnem >> 8) {
first = 1;
for (mnem = ph->mnemonic; (c = mnem & 0xff) != 0; mnem = mnem >> 8) {
if ((c == '/') && (option_phoneme_variants == 0))
break; // discard phoneme variant indicator

@@ -528,6 +536,7 @@ const char *GetTranslatedPhonemeString(int phoneme_mode)

*/

int ix;
unsigned int len;
int phon_out_ix = 0;
int stress;
@@ -562,7 +571,7 @@ const char *GetTranslatedPhonemeString(int phoneme_mode)
use_tie = 0;
}

for (int ix = 1; ix < (n_phoneme_list-2); ix++) {
for (ix = 1; ix < (n_phoneme_list-2); ix++) {
buf = phon_buf;

plist = &phoneme_list[ix];
@@ -710,12 +719,14 @@ int IsVowel(Translator *tr, int letter)

static int Unpronouncable2(Translator *tr, char *word)
{
int c;
int end_flags;
char ph_buf[N_WORD_PHONEMES];

ph_buf[0] = 0;
int c = word[-1];
c = word[-1];
word[-1] = ' '; // ensure there is a space before the "word"
int end_flags = TranslateRules(tr, word, ph_buf, sizeof(ph_buf), NULL, FLAG_UNPRON_TEST, NULL);
end_flags = TranslateRules(tr, word, ph_buf, sizeof(ph_buf), NULL, FLAG_UNPRON_TEST, NULL);
word[-1] = c;
if ((end_flags == 0) || (end_flags & SUFX_UNPRON))
return 1;
@@ -733,6 +744,8 @@ int Unpronouncable(Translator *tr, char *word, int posn)
int c;
int c1 = 0;
int vowel_posn = 9;
int index;
int count;
ALPHABET *alphabet;

utf8_in(&c, word);
@@ -752,8 +765,8 @@ int Unpronouncable(Translator *tr, char *word, int posn)
if (((c = *word) == ' ') || (c == 0) || (c == '\''))
return 0;

int index = 0;
int count = 0;
index = 0;
count = 0;
for (;;) {
index += utf8_in(&c, &word[index]);
if ((c == 0) || (c == ' '))
@@ -913,13 +926,14 @@ void ChangeWordStress(Translator *tr, char *word, int new_stress)
{
int ix;
unsigned char *p;
int max_stress;
int vowel_count; // num of vowels + 1
int stressed_syllable = 0; // position of stressed syllable
unsigned char phonetic[N_WORD_PHONEMES];
signed char vowel_stress[N_WORD_PHONEMES/2];

strcpy((char *)phonetic, word);
int max_stress = GetVowelStress(tr, phonetic, vowel_stress, &vowel_count, &stressed_syllable, 0);
max_stress = GetVowelStress(tr, phonetic, vowel_stress, &vowel_count, &stressed_syllable, 0);

if (new_stress >= 4) {
// promote to primary stress
@@ -987,6 +1001,7 @@ void SetWordStress(Translator *tr, char *output, unsigned int *dictionary_flags,
int mnem;
int opt_length;
int done;
int stressflags;
int dflags = 0;
int first_primary;
int long_vowel;
@@ -1005,7 +1020,7 @@ void SetWordStress(Translator *tr, char *output, unsigned int *dictionary_flags,
3 secondary stress
4 main stress */

int stressflags = tr->langopts.stress_flags;
stressflags = tr->langopts.stress_flags;

if (dictionary_flags != NULL)
dflags = dictionary_flags[0];
@@ -2343,12 +2358,14 @@ void ApplySpecialAttribute2(Translator *tr, char *phonemes, int dict_flags)
{
// apply after the translation is complete

int ix;
int len;
char *p;

int len = strlen(phonemes);
len = strlen(phonemes);

if (tr->langopts.param[LOPT_ALT] & 2) {
for (int ix = 0; ix < (len-1); ix++) {
for (ix = 0; ix < (len-1); ix++) {
if (phonemes[ix] == phonSTRESS_P) {
p = &phonemes[ix+1];
if ((dict_flags & FLAG_ALT2_TRANS) != 0) {
@@ -2376,23 +2393,28 @@ int TransposeAlphabet(Translator *tr, char *text)
int c;
int c2;
int ix;
int offset;
int min;
int max;
const char *map;
char *p = text;
char *p2;
int all_alpha = 1;
int bits;
int acc;
int pairs_start;
const short *pairs_list;
int bufix;
char buf[N_WORD_BYTES+1];

int offset = tr->transpose_min - 1;
int min = tr->transpose_min;
int max = tr->transpose_max;
offset = tr->transpose_min - 1;
min = tr->transpose_min;
max = tr->transpose_max;
map = tr->transpose_map;

int pairs_start = max - min + 2;
pairs_start = max - min + 2;

int bufix = 0;
bufix = 0;
do {
p += utf8_in(&c, p);
if (c != 0) {
@@ -2468,6 +2490,7 @@ static const char *LookupDict2(Translator *tr, const char *word, const char *wor
{
char *p;
char *next;
int hash;
int phoneme_len;
int wlen;
unsigned char flag;
@@ -2482,13 +2505,14 @@ static const char *LookupDict2(Translator *tr, const char *word, const char *wor
const char *word_end;
const char *word1;
int wflags = 0;
int lookup_symbol;
char word_buf[N_WORD_BYTES+1];
char dict_flags_buf[80];

if (wtab != NULL)
wflags = wtab->flags;

int lookup_symbol = flags[1] & FLAG_LOOKUP_SYMBOL;
lookup_symbol = flags[1] & FLAG_LOOKUP_SYMBOL;
word1 = word;
if (tr->transpose_min > 0) {
strncpy0(word_buf, word, N_WORD_BYTES);
@@ -2497,7 +2521,7 @@ static const char *LookupDict2(Translator *tr, const char *word, const char *wor
} else
wlen = strlen(word);

int hash = HashDictionary(word);
hash = HashDictionary(word);
p = tr->dict_hashtab[hash];

if (p == NULL) {
@@ -2740,6 +2764,7 @@ static const char *LookupDict2(Translator *tr, const char *word, const char *wor
*/
int LookupDictList(Translator *tr, char **wordptr, char *ph_out, unsigned int *flags, int end_flags, WORD_TAB *wtab)
{
int length;
const char *found;
const char *word1;
const char *word2;
@@ -2749,7 +2774,7 @@ int LookupDictList(Translator *tr, char **wordptr, char *ph_out, unsigned int *f
char word[N_WORD_BYTES];
static char word_replacement[N_WORD_BYTES];

int length = 0;
length = 0;
word2 = word1 = *wordptr;

while ((word2[nbytes = utf8_nbytes(word2)] == ' ') && (word2[nbytes+1] == '.')) {
@@ -2870,6 +2895,7 @@ int Lookup(Translator *tr, const char *word, char *ph_out)

int flags0;
unsigned int flags[2];
int say_as;
char *word1 = (char *)word;
char text[80];

@@ -2879,7 +2905,7 @@ int Lookup(Translator *tr, const char *word, char *ph_out)
flags0 = flags[0];

if (flags[0] & FLAG_TEXTMODE) {
int say_as = option_sayas;
say_as = option_sayas;
option_sayas = 0; // don't speak replacement word as letter names
text[0] = 0;
strncpy0(&text[1], word1, sizeof(text));
@@ -2913,7 +2939,10 @@ int RemoveEnding(Translator *tr, char *word, int end_type, char *word_copy)
that were done when the suffix was added to the original word.
*/

int i;
char *word_end;
int len_ending;
int end_flags;
const char *p;
int len;
char ending[50];
@@ -2932,7 +2961,7 @@ int RemoveEnding(Translator *tr, char *word, int end_type, char *word_copy)
if (*word_end == REPLACED_E)
*word_end = 'e';
}
int i = word_end - word;
i = word_end - word;

if (word_copy != NULL) {
memcpy(word_copy, word, i);
@@ -2940,7 +2969,6 @@ int RemoveEnding(Translator *tr, char *word, int end_type, char *word_copy)
}

// look for multibyte characters to increase the number of bytes to remove
int len_ending;
for (len_ending = i = (end_type & 0x3f); i > 0; i--) { // num.of characters of the suffix
word_end--;
while ((*word_end & 0xc0) == 0x80) {
@@ -2957,7 +2985,7 @@ int RemoveEnding(Translator *tr, char *word, int end_type, char *word_copy)
ending[i] = 0;
word_end--; // now pointing at last character of stem

int end_flags = (end_type & 0xfff0) | FLAG_SUFX;
end_flags = (end_type & 0xfff0) | FLAG_SUFX;

/* add an 'e' to the stem if appropriate,
if stem ends in vowel+consonant

+ 5
- 3
src/libespeak-ng/ieee80.c View File

@@ -49,13 +49,15 @@
double ConvertFromIeeeExtended(unsigned char *bytes /* LCN */)
{
double f;
int expon;
unsigned long hiMant, loMant;
int expon = ((bytes[0] & 0x7F) << 8) | (bytes[1] & 0xFF);
int hiMant = ((unsigned long)(bytes[2] & 0xFF) << 24)
expon = ((bytes[0] & 0x7F) << 8) | (bytes[1] & 0xFF);
hiMant = ((unsigned long)(bytes[2] & 0xFF) << 24)
| ((unsigned long)(bytes[3] & 0xFF) << 16)
| ((unsigned long)(bytes[4] & 0xFF) << 8)
| ((unsigned long)(bytes[5] & 0xFF));
int loMant = ((unsigned long)(bytes[6] & 0xFF) << 24)
loMant = ((unsigned long)(bytes[6] & 0xFF) << 24)
| ((unsigned long)(bytes[7] & 0xFF) << 16)
| ((unsigned long)(bytes[8] & 0xFF) << 8)
| ((unsigned long)(bytes[9] & 0xFF));

+ 43
- 25
src/libespeak-ng/intonation.c View File

@@ -320,6 +320,7 @@ static int no_tonic;

static void count_pitch_vowels(int start, int end, int clause_end)
{
int ix;
int stress;
int max_stress = 0;
int max_stress_posn = 0; // last syllable ot the highest stress
@@ -330,7 +331,7 @@ static void count_pitch_vowels(int start, int end, int clause_end)
number_tail = 0; // number between tonic syllable and next primary
last_primary = -1;

for (int ix = start; ix < end; ix++) {
for (ix = start; ix < end; ix++) {
stress = syllable_tab[ix].stress; // marked stress level

if (stress >= max_stress) {
@@ -387,18 +388,19 @@ static int count_increments(int ix, int end_ix, int min_stress)
// Set the pitch of a vowel in syllable_tab
static void set_pitch(SYLLABLE *syl, int base, int drop)
{
int pitch1, pitch2;
int flags = 0;

if (base < 0) base = 0;

int pitch2 = base;
pitch2 = base;

if (drop < 0) {
flags = SYL_RISE;
drop = -drop;
}

int pitch1 = pitch2 + drop;
pitch1 = pitch2 + drop;
if (pitch1 < 0)
pitch1 = 0;

@@ -412,7 +414,8 @@ static void set_pitch(SYLLABLE *syl, int base, int drop)

static int CountUnstressed(int start, int end, int limit)
{
int ix;
int ix;

for (ix = start; ix <= end; ix++) {
if (syllable_tab[ix].stress >= limit)
break;
@@ -429,7 +432,10 @@ static int SetHeadIntonation(TUNE *tune, int syl_ix, int end_ix, int control)
int increment = 0;
int n_steps = 0;
int stage; // onset, head, last
int initial;
int overflow_ix = 0;
int pitch_range;
int pitch_range_abs;
int *drops;
int n_unstressed = 0;
int unstressed_ix = 0;
@@ -438,10 +444,10 @@ static int SetHeadIntonation(TUNE *tune, int syl_ix, int end_ix, int control)
int head_final = end_ix;
int secondary = 2;

int pitch_range = (tune->head_end - tune->head_start) << 8;
int pitch_range_abs = abs(pitch_range);
pitch_range = (tune->head_end - tune->head_start) << 8;
pitch_range_abs = abs(pitch_range);
drops = drops_0; // this should be controled by tune->head_drops
int initial = 1;
initial = 1;

stage = 0;
if (tune->onset == 255)
@@ -545,6 +551,8 @@ static int calc_pitch_segment(int ix, int end_ix, TONE_HEAD *th, TONE_NUCLEUS *t
int initial;
int overflow = 0;
int n_overflow;
int pitch_range;
int pitch_range_abs;
int *drops;
signed char *overflow_tab;
SYLLABLE *syl;
@@ -552,8 +560,8 @@ static int calc_pitch_segment(int ix, int end_ix, TONE_HEAD *th, TONE_NUCLEUS *t
static signed char continue_tab[5] = { -26, 32, 20, 8, 0 };

drops = th->body_drops;
int pitch_range = (th->body_end - th->body_start) << 8;
int pitch_range_abs = abs(pitch_range);
pitch_range = (th->body_end - th->body_start) << 8;
pitch_range_abs = abs(pitch_range);

if (continuing) {
initial = 0;
@@ -630,12 +638,16 @@ static void SetPitchGradient(int start_ix, int end_ix, int start_pitch, int end_
// Set a linear pitch change over a number of syllables.
// Used for pre-head, unstressed syllables in the body, and the tail

int ix;
int stress;
int pitch;
int increment;
int n_increments;
int drop;
SYLLABLE *syl;

int increment = (end_pitch - start_pitch) << 8;
int n_increments = end_ix - start_ix;
increment = (end_pitch - start_pitch) << 8;
n_increments = end_ix - start_ix;

if (n_increments <= 0)
return;
@@ -643,9 +655,9 @@ static void SetPitchGradient(int start_ix, int end_ix, int start_pitch, int end_
if (n_increments > 1)
increment = increment / n_increments;

int pitch = start_pitch << 8;
pitch = start_pitch << 8;

for (int ix = start_ix; ix < end_ix; ix++) {
for (ix = start_ix; ix < end_ix; ix++) {
syl = &syllable_tab[ix];
stress = syl->stress;

@@ -669,10 +681,12 @@ static void SetPitchGradient(int start_ix, int end_ix, int start_pitch, int end_
// Calculate pitch values for the vowels in this tone group
static int calc_pitches2(int start, int end, int tune_number)
{
int ix;
TUNE *tune;
int drop;

tune = &tunes[tune_number];
int ix = start;
ix = start;

// vowels before the first primary stress

@@ -690,7 +704,6 @@ static int calc_pitches2(int start, int end, int tune_number)

// tonic syllable

int drop;
if (number_tail == 0) {
tone_pitch_env = tune->nucleus0_env;
drop = tune->nucleus0_max - tune->nucleus0_min;
@@ -715,8 +728,10 @@ static int calc_pitches2(int start, int end, int tune_number)
// Calculate pitch values for the vowels in this tone group
static int calc_pitches(int control, int start, int end, int tune_number)
{
int ix;
TONE_HEAD *th;
TONE_NUCLEUS *tn;
int drop;
int continuing = 0;

if (control == 0)
@@ -727,7 +742,7 @@ static int calc_pitches(int control, int start, int end, int tune_number)

th = &tone_head_table[tune_number];
tn = &tone_nucleus_table[tune_number];
int ix = start;
ix = start;

// vowels before the first primary stress

@@ -748,7 +763,6 @@ static int calc_pitches(int control, int start, int end, int tune_number)
if (tn->flags & T_EMPH)
syllable_tab[ix].flags |= SYL_EMPHASIS;

int drop;
if (number_tail == 0) {
tone_pitch_env = tn->pitch_env0;
drop = tn->tonic_max0 - tn->tonic_min0;
@@ -775,6 +789,7 @@ static void CalcPitches_Tone(Translator *tr, int clause_tone)
// clause_tone: 0=. 1=, 2=?, 3=! 4=none

PHONEME_LIST *p;
int ix;
int count_stressed = 0;
int final_stressed = 0;

@@ -793,7 +808,7 @@ static void CalcPitches_Tone(Translator *tr, int clause_tone)

// count number of stressed syllables
p = &phoneme_list[0];
for (int ix = 0; ix < n_phoneme_list; ix++, p++) {
for (ix = 0; ix < n_phoneme_list; ix++, p++) {
if ((p->type == phVOWEL) && (p->stresslevel >= 4)) {
if (count_stressed == 0)
final_stressed = ix;
@@ -822,7 +837,7 @@ static void CalcPitches_Tone(Translator *tr, int clause_tone)
prev_tph = prevw_tph = phoneme_tab[phonPAUSE];

// perform tone sandhi
for (int ix = 0; ix < n_phoneme_list; ix++, p++) {
for (ix = 0; ix < n_phoneme_list; ix++, p++) {
if ((p->type == phPAUSE) && (p->ph->std_length > 50)) {
pause = 1; // there is a pause since the previous vowel
prevw_tph = phoneme_tab[phonPAUSE]; // forget previous tone
@@ -887,7 +902,7 @@ static void CalcPitches_Tone(Translator *tr, int clause_tone)

// convert tone numbers to pitch
p = &phoneme_list[0];
for (int ix = 0; ix < n_phoneme_list; ix++, p++) {
for (ix = 0; ix < n_phoneme_list; ix++, p++) {
if (p->synthflags & SFLAG_SYLLABLE) {
tone_ph = p->tone_ph;

@@ -921,21 +936,24 @@ void CalcPitches(Translator *tr, int clause_type)
int ix;
int x;
int st_ix;
int n_st;
int option;
int group_tone;
int group_tone_comma;
int ph_start = 0;
int st_start;
int st_clause_end;
int count;
int n_primary;
int count_primary;
PHONEME_TAB *ph;
int ph_end = n_phoneme_list;

SYLLABLE syllable_tab2[N_PHONEME_LIST];

syllable_tab = syllable_tab2; // don't use permanent storage. it's only needed during the call of CalcPitches()
int n_st = 0;
int n_primary = 0;

n_st = 0;
n_primary = 0;
for (ix = 0; ix < (n_phoneme_list-1); ix++) {
p = &phoneme_list[ix];
if (p->synthflags & SFLAG_SYLLABLE) {
@@ -976,8 +994,8 @@ void CalcPitches(Translator *tr, int clause_type)
else
no_tonic = 0;

int st_start = 0;
int count_primary = 0;
st_start = 0;
count_primary = 0;
for (st_ix = 0; st_ix < n_st; st_ix++) {
syl = &syllable_tab[st_ix];


+ 45
- 21
src/libespeak-ng/klatt.c View File

@@ -127,7 +127,9 @@ static short natural_samples[100] = {

static double resonator(resonator_ptr r, double input)
{
double x = (double)((double)r->a * (double)input + (double)r->b * (double)r->p1 + (double)r->c * (double)r->p2);
double x;

x = (double)((double)r->a * (double)input + (double)r->b * (double)r->p1 + (double)r->c * (double)r->p2);
r->p2 = (double)r->p1;
r->p1 = (double)x;

@@ -136,7 +138,9 @@ static double resonator(resonator_ptr r, double input)

static double resonator2(resonator_ptr r, double input)
{
double x = (double)((double)r->a * (double)input + (double)r->b * (double)r->p1 + (double)r->c * (double)r->p2);
double x;

x = (double)((double)r->a * (double)input + (double)r->b * (double)r->p1 + (double)r->c * (double)r->p2);
r->p2 = (double)r->p1;
r->p1 = (double)x;

@@ -173,12 +177,15 @@ static double antiresonator2(resonator_ptr r, double input)
static void flutter(klatt_frame_ptr frame)
{
static int time_count;
double fla = (double)kt_globals.f0_flutter / 50;
double flb = (double)kt_globals.original_f0 / 100;
double flc = sin(PI*12.7*time_count); // because we are calling flutter() more frequently, every 2.9mS
double fld = sin(PI*7.1*time_count);
double fle = sin(PI*4.7*time_count);
double delta_f0 = fla * flb * (flc + fld + fle) * 10;
double delta_f0;
double fla, flb, flc, fld, fle;

fla = (double)kt_globals.f0_flutter / 50;
flb = (double)kt_globals.original_f0 / 100;
flc = sin(PI*12.7*time_count); // because we are calling flutter() more frequently, every 2.9mS
fld = sin(PI*7.1*time_count);
fle = sin(PI*4.7*time_count);
delta_f0 = fla * flb * (flc + fld + fle) * 10;
frame->F0hz10 = frame->F0hz10 + (long)delta_f0;
time_count++;
}
@@ -192,7 +199,13 @@ static void flutter(klatt_frame_ptr frame)

static double sampled_source(int source_num)
{
int itemp;
double ftemp;
double result;
double diff_value;
int current_value;
int next_value;
double temp_diff;
short *samples;

if (source_num == 0) {
@@ -204,17 +217,17 @@ static double sampled_source(int source_num)
}

if (kt_globals.T0 != 0) {
double ftemp = (double)kt_globals.nper;
ftemp = (double)kt_globals.nper;
ftemp = ftemp / kt_globals.T0;
ftemp = ftemp * kt_globals.num_samples;
int itemp = (int)ftemp;
itemp = (int)ftemp;

double temp_diff = ftemp - (double)itemp;
temp_diff = ftemp - (double)itemp;

int current_value = samples[itemp];
int next_value = samples[itemp+1];
current_value = samples[itemp];
next_value = samples[itemp+1];

double diff_value = (double)next_value - (double)current_value;
diff_value = (double)next_value - (double)current_value;
diff_value = diff_value * temp_diff;

result = samples[itemp] + diff_value;
@@ -546,12 +559,13 @@ static double impulsive_source()

static double natural_source()
{
double lgtemp;
static double vwave;

if (kt_globals.nper < kt_globals.nopen) {
kt_globals.pulse_shape_a -= kt_globals.pulse_shape_b;
vwave += kt_globals.pulse_shape_a;
double lgtemp = vwave * 0.028;
lgtemp = vwave * 0.028;

return lgtemp;
}
@@ -707,9 +721,12 @@ static void pitch_synch_par_reset(klatt_frame_ptr frame)

static void setabc(long int f, long int bw, resonator_ptr rp)
{
double r;
double arg;

// Let r = exp(-pi bw t)
double arg = kt_globals.minus_pi_t * bw;
double r = exp(arg);
arg = kt_globals.minus_pi_t * bw;
r = exp(arg);

// Let c = -r**2
rp->c = -(r * r);
@@ -731,12 +748,15 @@ static void setabc(long int f, long int bw, resonator_ptr rp)

static void setzeroabc(long int f, long int bw, resonator_ptr rp)
{
double r;
double arg;

f = -f;

// First compute ordinary resonator coefficients
// Let r = exp(-pi bw t)
double arg = kt_globals.minus_pi_t * bw;
double r = exp(arg);
arg = kt_globals.minus_pi_t * bw;
r = exp(arg);

// Let c = -r**2
rp->c = -(r * r);
@@ -771,9 +791,10 @@ static void setzeroabc(long int f, long int bw, resonator_ptr rp)

static double gen_noise(double noise)
{
long temp;
static double nlast;

long temp = (long)getrandom(-8191, 8191);
temp = (long)getrandom(-8191, 8191);
kt_globals.nrand = (long)temp;

noise = kt_globals.nrand + (0.75 * nlast);
@@ -1042,11 +1063,14 @@ int Wavegen_Klatt2(int length, int modulation, int resume, frame_t *fr1, frame_t

void KlattInit()
{

static short formant_hz[10] = { 280, 688, 1064, 2806, 3260, 3700, 6500, 7000, 8000, 280 };
static short bandwidth[10] = { 89, 160, 70, 160, 200, 200, 500, 500, 500, 89 };
static short parallel_amp[10] = { 0, 59, 59, 59, 59, 59, 59, 0, 0, 0 };
static short parallel_bw[10] = { 59, 59, 89, 149, 200, 200, 500, 0, 0, 0 };

int ix;

sample_count = 0;

kt_globals.synthesis_model = CASCADE_PARALLEL;
@@ -1064,7 +1088,7 @@ void KlattInit()
KlattReset(2);

// set default values for frame parameters
for (int ix = 0; ix <= 9; ix++) {
for (ix = 0; ix <= 9; ix++) {
kt_frame.Fhz[ix] = formant_hz[ix];
kt_frame.Bhz[ix] = bandwidth[ix];
kt_frame.Ap[ix] = parallel_amp[ix];

+ 17
- 11
src/libespeak-ng/mbrowrap.c View File

@@ -123,7 +123,7 @@ static void close_pipes(int p1[2], int p2[2], int p3[2])

static int start_mbrola(const char *voice_path)
{
int p_stdin[2], p_stdout[2], p_stderr[2];
int error, p_stdin[2], p_stdout[2], p_stderr[2];
ssize_t written;
char charbuf[20];

@@ -132,7 +132,7 @@ static int start_mbrola(const char *voice_path)
return -1;
}

int error = create_pipes(p_stdin, p_stdout, p_stderr);
error = create_pipes(p_stdin, p_stdout, p_stderr);
if (error)
return -1;

@@ -146,6 +146,8 @@ static int start_mbrola(const char *voice_path)
}

if (mbr_pid == 0) {
int i;

if (dup2(p_stdin[0], 0) == -1 ||
dup2(p_stdout[1], 1) == -1 ||
dup2(p_stderr[1], 2) == -1) {
@@ -155,7 +157,8 @@ static int start_mbrola(const char *voice_path)
(void)written; // suppress 'variable not used' warning
_exit(1);
}
for (int i = p_stderr[1]; i > 2; i--)

for (i = p_stderr[1]; i > 2; i--)
close(i);
signal(SIGHUP, SIG_IGN);
signal(SIGINT, SIG_IGN);
@@ -239,7 +242,7 @@ static void free_pending_data(void)
static int mbrola_died(void)
{
pid_t pid;
int status;
int status, len;
const char *msg;
char msgbuf[80];

@@ -266,7 +269,7 @@ static int mbrola_died(void)

log("mbrowrap error: %s", msg);

size_t len = strlen(mbr_errorbuf);
len = strlen(mbr_errorbuf);
if (!len)
snprintf(mbr_errorbuf, sizeof(mbr_errorbuf), "%s", msg);
else
@@ -323,11 +326,12 @@ static int mbrola_has_errors(void)
static int send_to_mbrola(const char *cmd)
{
ssize_t result;
int len;

if (!mbr_pid)
return -1;

int len = strlen(cmd);
len = strlen(cmd);
result = write(mbr_cmd_fd, cmd, len);

if (result == -1) {
@@ -477,13 +481,14 @@ static ssize_t receive_from_mbrola(void *buffer, size_t bufsize)

int init_MBR(const char *voice_path)
{
int error, result;
unsigned char wavhdr[45];

int error = start_mbrola(voice_path);
error = start_mbrola(voice_path);
if (error)
return -1;

int result = send_to_mbrola("#\n");
result = send_to_mbrola("#\n");
if (result != 2) {
stop_mbrola();
return -1;
@@ -528,7 +533,7 @@ void close_MBR(void)

int reset_MBR()
{
int success = 1;
int result, success = 1;
char dummybuf[4096];

if (mbr_state == MBR_IDLE)
@@ -538,7 +543,7 @@ int reset_MBR()
if (kill(mbr_pid, SIGUSR1) == -1)
success = 0;
free_pending_data();
int result = write(mbr_cmd_fd, "\n#\n", 3);
result = write(mbr_cmd_fd, "\n#\n", 3);
if (result != 3)
success = 0;
do {
@@ -592,9 +597,10 @@ void setVolumeRatio_MBR(float value)

int lastErrorStr_MBR(char *buffer, int bufsize)
{
int result;
if (mbr_pid)
mbrola_has_errors();
int result = snprintf(buffer, bufsize, "%s", mbr_errorbuf);
result = snprintf(buffer, bufsize, "%s", mbr_errorbuf);
return result >= bufsize ? (bufsize - 1) : result;
}


+ 45
- 22
src/libespeak-ng/numbers.c View File

@@ -390,11 +390,12 @@ static const unsigned short letter_accents_250[] = {

static int LookupLetter2(Translator *tr, unsigned int letter, char *ph_buf)
{
int len;
char single_letter[10];

single_letter[0] = 0;
single_letter[1] = '_';
int len = utf8_out(letter, &single_letter[2]);
len = utf8_out(letter, &single_letter[2]);
single_letter[len+2] = ' ';
single_letter[len+3] = 0;

@@ -477,12 +478,13 @@ void LookupLetter(Translator *tr, unsigned int letter, int next_byte, char *ph_b
{
// control, bit 0: not the first letter of a word

int len;
static char single_letter[10] = { 0, 0 };
unsigned int dict_flags[2];
char ph_buf3[40];

ph_buf1[0] = 0;
int len = utf8_out(letter, &single_letter[2]);
len = utf8_out(letter, &single_letter[2]);
single_letter[len+2] = ' ';

if (next_byte == -1) {
@@ -649,8 +651,10 @@ static const char *hex_letters[] = {
int IsSuperscript(int letter)
{
// is this a subscript or superscript letter ?
int ix;
int c;
for (int ix = 0; (c = derived_letters[ix]) != 0; ix += 2) {

for (ix = 0; (c = derived_letters[ix]) != 0; ix += 2) {
if (c > letter)
break;
if (c == letter)
@@ -667,6 +671,7 @@ int TranslateLetter(Translator *tr, char *word, char *phonemes, int control)
// bit 1: say 'capital'
// bit 2: say character code for unknown letters

int n_bytes;
int letter;
int len;
int ix;
@@ -675,8 +680,11 @@ int TranslateLetter(Translator *tr, char *word, char *phonemes, int control)
char *pbuf;
const char *modifier;
ALPHABET *alphabet;
int al_offset;
int al_flags;
int language;
int number;
int phontab_1;
int speak_letter_number;
char capital[30];
char ph_buf[80];
@@ -688,9 +696,9 @@ int TranslateLetter(Translator *tr, char *word, char *phonemes, int control)
ph_buf[0] = 0;
ph_alphabet[0] = 0;
capital[0] = 0;
int phontab_1 = translator->phoneme_tab_ix;
phontab_1 = translator->phoneme_tab_ix;

int n_bytes = utf8_in(&letter, word);
n_bytes = utf8_in(&letter, word);

if ((letter & 0xfff00) == 0x0e000)
letter &= 0xff; // uncode private usage area
@@ -738,8 +746,8 @@ int TranslateLetter(Translator *tr, char *word, char *phonemes, int control)
LookupLetter(tr, number, 0, ph_buf, control & 1);
}

int al_offset = 0;
int al_flags = 0;
al_offset = 0;
al_flags = 0;
if ((alphabet = AlphabetFromChar(letter)) != NULL) {
al_offset = alphabet->offset;
al_flags = alphabet->flags;
@@ -903,6 +911,7 @@ void SetSpellingStress(Translator *tr, char *phonemes, int control, int n_chars)
unsigned int c;
int n_stress = 0;
int prev = 0;
int count;
unsigned char buf[N_WORD_PHONEMES];

for (ix = 0; (c = phonemes[ix]) != 0; ix++) {
@@ -912,7 +921,7 @@ void SetSpellingStress(Translator *tr, char *phonemes, int control, int n_chars)
}
buf[ix] = 0;

int count = 0;
count = 0;
prev = 0;
for (ix = 0; (c = buf[ix]) != 0; ix++) {
if ((c == phonSTRESS_P) && (n_chars > 1) && (prev != phonSWITCH)) {
@@ -1019,7 +1028,10 @@ int TranslateRoman(Translator *tr, char *word, char *ph_out, WORD_TAB *wtab)
int c;
char *p;
const char *p2;
int acc;
int prev;
int value;
int subtract;
int repeat = 0;
int n_digits = 0;
char *word_start;
@@ -1031,9 +1043,9 @@ int TranslateRoman(Translator *tr, char *word, char *ph_out, WORD_TAB *wtab)
static const char *roman_numbers = "ixcmvld";
static int roman_values[] = { 1, 10, 100, 1000, 5, 50, 500 };

int acc = 0;
int prev = 0;
int subtract = 0x7fff;
acc = 0;
prev = 0;
subtract = 0x7fff;
ph_out[0] = 0;
flags[0] = 0;
flags[1] = 0;
@@ -1271,7 +1283,11 @@ static int LookupNum2(Translator *tr, int value, int thousandplex, const int con
// bit 8 followed by decimal fraction
// bit 9: use #f form for both tens and units (lang=ml)

int found;
int ix;
int units;
int tens;
int is_ordinal;
int used_and = 0;
int found_ordinal = 0;
int next_phtype;
@@ -1282,10 +1298,10 @@ static int LookupNum2(Translator *tr, int value, int thousandplex, const int con
char ph_digits[50];
char ph_and[12];

int units = value % 10;
int tens = value / 10;
units = value % 10;
tens = value / 10;

int found = 0;
found = 0;
ph_ordinal[0] = 0;
ph_tens[0] = 0;
ph_digits[0] = 0;
@@ -1294,7 +1310,7 @@ static int LookupNum2(Translator *tr, int value, int thousandplex, const int con
if (control & 0x20)
ord_type = 'q';

int is_ordinal = control & 1;
is_ordinal = control & 1;

if ((control & 2) && (n_digit_lookup == 2)) {
// pronunciation of the final 2 digits has already been found
@@ -1508,8 +1524,15 @@ static int LookupNum3(Translator *tr, int value, char *ph_out, int suppress_null
// bit 8 followed by decimal fraction

int found;
int hundreds;
int tensunits;
int x;
int ix;
int exact;
int ordinal;
int tplex;
int say_zero_hundred = 0;
int say_one_hundred;
char string[12]; // for looking up entries in **_list
char buf1[100];
char buf2[100];
@@ -1520,9 +1543,9 @@ static int LookupNum3(Translator *tr, int value, char *ph_out, int suppress_null
char ph_hundred_and[12];
char ph_thousand_and[12];

int ordinal = control & 0x22;
int hundreds = value / 100;
int tensunits = value % 100;
ordinal = control & 0x22;
hundreds = value / 100;
tensunits = value % 100;
buf1[0] = 0;

ph_thousands[0] = 0;
@@ -1551,11 +1574,11 @@ static int LookupNum3(Translator *tr, int value, char *ph_out, int suppress_null
} else if (hundreds >= 10) {
ph_digits[0] = 0;

int exact = 0;
exact = 0;
if ((value % 1000) == 0)
exact = 1;

int tplex = thousandplex+1;
tplex = thousandplex+1;
if (tr->langopts.numbers2 & NUM2_MYRIADS)
tplex = 0;

@@ -1623,7 +1646,7 @@ static int LookupNum3(Translator *tr, int value, char *ph_out, int suppress_null
if (found)
ph_100[0] = 0;
else {
int say_one_hundred = 1;
say_one_hundred = 1;
if (hundreds == 1) {
if ((tr->langopts.numbers & NUM_OMIT_1_HUNDRED) != 0)
say_one_hundred = 0;
@@ -1681,7 +1704,7 @@ static int LookupNum3(Translator *tr, int value, char *ph_out, int suppress_null
}
} else {
if (ph_ordinal2[0] != 0) {
int ix = strlen(buf1);
ix = strlen(buf1);
if ((ix > 0) && (buf1[ix-1] == phonPAUSE_SHORT))
buf1[ix-1] = 0; // remove pause before addding ordinal suffix
strcpy(buf2, ph_ordinal2);

+ 14
- 8
src/libespeak-ng/phonemelist.c View File

@@ -43,7 +43,7 @@ static int SubstitutePhonemes(Translator *tr, PHONEME_LIST *plist_out)
{
// Copy the phonemes list and perform any substitutions that are required for the
// current voice
int ix;
int k;
int replace_flags;
int n_plist_out = 0;
@@ -51,7 +51,7 @@ static int SubstitutePhonemes(Translator *tr, PHONEME_LIST *plist_out)
PHONEME_LIST2 *plist2;
PHONEME_TAB *next = NULL;

for (int ix = 0; (ix < n_ph_list2) && (n_plist_out < N_PHONEME_LIST); ix++) {
for (ix = 0; (ix < n_ph_list2) && (n_plist_out < N_PHONEME_LIST); ix++) {
plist2 = &ph_list2[ix];

// don't do any substitution if the language has been temporarily changed
@@ -108,13 +108,19 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
PHONEME_TAB *next, *next2;
int unstress_count = 0;
int word_stress = 0;
int current_phoneme_tab;
int max_stress;
int voicing;
int regression;
int end_sourceix;
int alternative;
int delete_count;
int word_start;
int inserted;
int deleted;
PHONEME_DATA phdata;

int n_ph_list3;
PHONEME_LIST *plist3;
PHONEME_LIST *plist3_inserted = NULL;
PHONEME_LIST ph_list3[N_PHONEME_LIST];
@@ -125,10 +131,10 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
memset(&worddata, 0, sizeof(worddata));
plist2 = ph_list2;
phlist = phoneme_list;
int end_sourceix = plist2[n_ph_list2-1].sourceix;
end_sourceix = plist2[n_ph_list2-1].sourceix;

// is the last word of the clause unstressed ?
int max_stress = 0;
max_stress = 0;
for (j = n_ph_list2-3; j >= 0; j--) {
// start with the last phoneme (before the terminating pauses) and move backwards
if ((plist2[j].stresslevel & 0x7f) > max_stress)
@@ -151,8 +157,8 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
}

// look for switch of phoneme tables
int delete_count = 0;
int current_phoneme_tab = tr->phoneme_tab_ix;
delete_count = 0;
current_phoneme_tab = tr->phoneme_tab_ix;
for (j = 0; j < n_ph_list2; j++) {
if (current_phoneme_tab != tr->phoneme_tab_ix)
plist2[j].synthflags |= SFLAG_SWITCHED_LANG;
@@ -244,7 +250,7 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
}
}

int n_ph_list3 = SubstitutePhonemes(tr, ph_list3) - 2;
n_ph_list3 = SubstitutePhonemes(tr, ph_list3) - 2;

for (j = 0; (j < n_ph_list3) && (ix < N_PHONEME_LIST-3);) {
if (ph_list3[j].sourceix) {
@@ -272,7 +278,7 @@ void MakePhonemeList(Translator *tr, int post_pause, int start_sentence)
// transfer all the phonemes of the clause into phoneme_list
ph = phoneme_tab[phonPAUSE];
ph_list3[0].ph = ph;
int word_start = 1;
word_start = 1;

for (j = 0; insert_ph || ((j < n_ph_list3) && (ix < N_PHONEME_LIST-3)); j++) {
plist3 = &ph_list3[j];

+ 53
- 27
src/libespeak-ng/readclause.c View File

@@ -321,6 +321,7 @@ int iswalnum(int c)
int towlower(int c)
{
int x;
int ix;

if (c < 0x80)
return tolower(c);
@@ -330,7 +331,7 @@ int towlower(int c)

if (x == 0xfd) {
// special cases, lookup translation table
for (int ix = 0; wchar_tolower[ix] != 0; ix += 2) {
for (ix = 0; wchar_tolower[ix] != 0; ix += 2) {
if (wchar_tolower[ix] == c)
return wchar_tolower[ix+1];
}
@@ -340,12 +341,13 @@ int towlower(int c)

int towupper(int c)
{
int ix;
// check whether a previous character code is the upper-case equivalent of this character
if (towlower(c-32) == c)
return c-32; // yes, use it
if (towlower(c-1) == c)
return c-1;
for (int ix = 0; wchar_toupper[ix] != 0; ix += 2) {
for (ix = 0; wchar_toupper[ix] != 0; ix += 2) {
if (wchar_toupper[ix] == c)
return wchar_toupper[ix+1];
}
@@ -463,6 +465,7 @@ int iswupper2(int c)
int towlower2(unsigned int c)
{
int x;
int ix;

// check for non-standard upper to lower case conversions
if (c == 'I') {
@@ -481,7 +484,7 @@ int towlower2(unsigned int c)

if (x == 0xfd) {
// special cases, lookup translation table
for (int ix = 0; wchar_tolower[ix] != 0; ix += 2) {
for (ix = 0; wchar_tolower[ix] != 0; ix += 2) {
if (wchar_tolower[ix] == (int)c)
return wchar_tolower[ix+1];
}
@@ -491,6 +494,7 @@ int towlower2(unsigned int c)

int towupper2(unsigned int c)
{
int ix;
if (c > MAX_WALPHA)
return towupper(c);

@@ -499,7 +503,7 @@ int towupper2(unsigned int c)
return c-32; // yes, use it
if (towlower2(c-1) == (int)c)
return c-1;
for (int ix = 0; wchar_toupper[ix] != 0; ix += 2) {
for (ix = 0; wchar_toupper[ix] != 0; ix += 2) {
if (wchar_toupper[ix] == (int)c)
return wchar_toupper[ix+1];
}
@@ -539,13 +543,14 @@ int Eof(void)
static int GetC_get(void)
{
unsigned int c;
unsigned int c2;

if (f_input != NULL) {
c = fgetc(f_input);
if (feof(f_input)) c = ' ';

if (option_multibyte == espeakCHARS_16BIT) {
unsigned int c2 = fgetc(f_input);
c2 = fgetc(f_input);
if (feof(f_input)) c2 = 0;
c = c + (c2 << 8);
}
@@ -583,10 +588,12 @@ static int GetC(void)
// Returns a unicode wide character
// Performs UTF8 checking and conversion

int c;
int c1;
int c2;
int cbuf[4];
int ix;
int n_bytes;
static int ungot2 = 0;
static const unsigned char mask[4] = { 0xff, 0x1f, 0x0f, 0x07 };

@@ -608,7 +615,7 @@ static int GetC(void)

if ((option_multibyte < 2) && (c1 & 0x80)) {
// multi-byte utf8 encoding, convert to unicode
int n_bytes = 0;
n_bytes = 0;

if (((c1 & 0xe0) == 0xc0) && ((c1 & 0x1e) != 0))
n_bytes = 1;
@@ -618,7 +625,7 @@ static int GetC(void)
n_bytes = 3;

if ((ix = n_bytes) > 0) {
int c = c1 & mask[ix];
c = c1 & mask[ix];
while (ix > 0) {
if ((c2 = cbuf[ix] = GetC_get()) == 0) {
if (option_multibyte == espeakCHARS_AUTO)
@@ -662,12 +669,12 @@ static void UngetC(int c)
const char *WordToString2(unsigned int word)
{
// Convert a language mnemonic word into a string
int ix;
static char buf[5];
char *p;

p = buf;
for (int ix = 3; ix >= 0; ix--) {
for (ix = 3; ix >= 0; ix--) {
if ((*p = word >> (ix*8)) != 0)
p++;
}
@@ -697,6 +704,7 @@ static const char *LookupCharName(Translator *tr, int c, int only)
// Find the phoneme string (in ascii) to speak the name of character c
// Used for punctuation characters and symbols

int ix;
unsigned int flags[2];
char single_letter[24];
char phonemes[60];
@@ -710,7 +718,7 @@ static const char *LookupCharName(Translator *tr, int c, int only)
flags[1] = 0;
single_letter[0] = 0;
single_letter[1] = '_';
int ix = utf8_out(c, &single_letter[2]);
ix = utf8_out(c, &single_letter[2]);
single_letter[2+ix] = 0;

if (only) {
@@ -780,6 +788,7 @@ static int LoadSoundFile(const char *fname, int index)
FILE *f;
char *p;
int *ip;
int length;
char fname_temp[100];
char fname2[sizeof(path_home)+13+40];

@@ -800,12 +809,13 @@ static int LoadSoundFile(const char *fname, int index)
f = NULL;
#ifdef PLATFORM_POSIX
if ((f = fopen(fname, "rb")) != NULL) {
int ix;
int fd_temp;
int header[3];
char command[sizeof(fname2)+sizeof(fname2)+40];

fseek(f, 20, SEEK_SET);
for (int ix = 0; ix < 3; ix++)
for (ix = 0; ix < 3; ix++)
header[ix] = Read4Bytes(f);

// if the sound file is not mono, 16 bit signed, at the correct sample rate, then convert it
@@ -832,7 +842,7 @@ static int LoadSoundFile(const char *fname, int index)
}
}

int length = GetFileLength(fname);
length = GetFileLength(fname);
fseek(f, 0, SEEK_SET);
if ((p = (char *)realloc(soundicon_tab[index].data, length)) == NULL) {
fclose(f);
@@ -851,7 +861,9 @@ static int LoadSoundFile(const char *fname, int index)
static int LookupSoundicon(int c)
{
// Find the sound icon number for a punctuation chatacter
for (int ix = N_SOUNDICON_SLOTS; ix < n_soundicon_tab; ix++) {
int ix;

for (ix = N_SOUNDICON_SLOTS; ix < n_soundicon_tab; ix++) {
if (soundicon_tab[ix].name == c) {
if (soundicon_tab[ix].length == 0) {
if (LoadSoundFile(NULL, ix) != 0)
@@ -868,7 +880,7 @@ static int LoadSoundFile2(const char *fname)
// Load a sound file into one of the reserved slots in the sound icon table
// (if it'snot already loaded)

int ix;
int ix;
static int slot = -1;

for (ix = 0; ix < n_soundicon_tab; ix++) {
@@ -900,11 +912,14 @@ static int AnnouncePunctuation(Translator *tr, int c1, int *c2_ptr, char *output
int soundicon;
int attributes;
int short_pause;
int c2;
int len;
int bufix1;
char buf[200];
char buf2[80];
char ph_buf[30];

int c2 = *c2_ptr;
c2 = *c2_ptr;
buf[0] = 0;

if ((soundicon = LookupSoundicon(c1)) >= 0) {
@@ -964,8 +979,8 @@ static int AnnouncePunctuation(Translator *tr, int c1, int *c2_ptr, char *output
}
}

int bufix1 = *bufix;
int len = strlen(buf);
bufix1 = *bufix;
len = strlen(buf);
strcpy(&output[*bufix], buf);
*bufix += len;

@@ -1057,6 +1072,7 @@ static const char *VoiceFromStack()
// Use the voice properties from the SSML stack to choose a voice, and switch
// to that voice if it's not the current voice

int ix;
const char *p;
SSML_STACK *sp;
const char *v_id;
@@ -1074,7 +1090,7 @@ static const char *VoiceFromStack()
voice_select.variant = ssml_stack[0].voice_variant_number;
voice_select.identifier = NULL;

for (int ix = 0; ix < n_ssml_stack; ix++) {
for (ix = 0; ix < n_ssml_stack; ix++) {
sp = &ssml_stack[ix];
voice_name_specified = 0;

@@ -1131,6 +1147,7 @@ static void ProcessParamStack(char *outbuf, int *outix)
{
// Set the speech parameters from the parameter stack
int param;
int ix;
int value;
char buf[20];
int new_parameters[N_SPEECH_PARAM];
@@ -1139,7 +1156,7 @@ static void ProcessParamStack(char *outbuf, int *outix)
for (param = 0; param < N_SPEECH_PARAM; param++)
new_parameters[param] = -1;

for (int ix = 0; ix < n_param_stack; ix++) {
for (ix = 0; ix < n_param_stack; ix++) {
for (param = 0; param < N_SPEECH_PARAM; param++) {
if (param_stack[ix].parameter[param] >= 0)
new_parameters[param] = param_stack[ix].parameter[param];
@@ -1176,6 +1193,7 @@ static void ProcessParamStack(char *outbuf, int *outix)

static PARAM_STACK *PushParamStack(int tag_type)
{
int ix;
PARAM_STACK *sp;

sp = &param_stack[n_param_stack];
@@ -1183,7 +1201,7 @@ static PARAM_STACK *PushParamStack(int tag_type)
n_param_stack++;

sp->type = tag_type;
for (int ix = 0; ix < N_SPEECH_PARAM; ix++)
for (ix = 0; ix < N_SPEECH_PARAM; ix++)
sp->parameter[ix] = -1;
return sp;
}
@@ -1191,12 +1209,13 @@ static PARAM_STACK *PushParamStack(int tag_type)
static void PopParamStack(int tag_type, char *outbuf, int *outix)
{
// unwind the stack up to and including the previous tag of this type
int ix;
int top = 0;

if (tag_type >= SSML_CLOSE)
tag_type -= SSML_CLOSE;

for (int ix = 0; ix < n_param_stack; ix++) {
for (ix = 0; ix < n_param_stack; ix++) {
if (param_stack[ix].type == tag_type)
top = ix;
}
@@ -1238,7 +1257,8 @@ static wchar_t *GetSsmlAttribute(wchar_t *pw, const char *name)

static int attrcmp(const wchar_t *string1, const char *string2)
{
int ix;
int ix;

if (string1 == NULL)
return 1;

@@ -1301,6 +1321,7 @@ static int attr_prosody_value(int param_type, const wchar_t *pw, int *value_out)
{
int sign = 0;
wchar_t *tail;
double value;

while (iswspace(*pw)) pw++;
if (*pw == '+') {
@@ -1311,7 +1332,7 @@ static int attr_prosody_value(int param_type, const wchar_t *pw, int *value_out)
pw++;
sign = -1;
}
double value = (double)wcstod(pw, &tail);
value = (double)wcstod(pw, &tail);
if (tail == pw) {
// failed to find a number, return 100%
*value_out = 100;
@@ -1326,8 +1347,9 @@ static int attr_prosody_value(int param_type, const wchar_t *pw, int *value_out)
}

if ((tail[0] == 's') && (tail[1] == 't')) {
double x;
// convert from semitones to a frequency percentage
double x = pow((double)2.0, (double)((value*sign)/12)) * 100;
x = pow((double)2.0, (double)((value*sign)/12)) * 100;
*value_out = (int)x;
return 2; // percentage
}
@@ -1469,6 +1491,7 @@ static int GetVoiceAttributes(wchar_t *pw, int tag_type)
static void SetProsodyParameter(int param_type, wchar_t *attr1, PARAM_STACK *sp)
{
int value;
int sign;

static const MNEM_TAB mnem_volume[] = {
{ "default", 100 },
@@ -1519,7 +1542,7 @@ static void SetProsodyParameter(int param_type, wchar_t *attr1, PARAM_STACK *sp)
// mnemonic specifies a value as a percentage of the base pitch/range/rate/volume
sp->parameter[param_type] = (param_stack[0].parameter[param_type] * value)/100;
} else {
int sign = attr_prosody_value(param_type, attr1, &value);
sign = attr_prosody_value(param_type, attr1, &value);

if (sign == 0)
sp->parameter[param_type] = value; // absolute value in Hz
@@ -1544,13 +1567,14 @@ static int ReplaceKeyName(char *outbuf, int index, int *outix)
{ NULL, 0 }
};

int ix;
int letter;
char *p;

p = &outbuf[index];

if ((letter = LookupMnem(keynames, p)) != 0) {
int ix = utf8_out(letter, p);
ix = utf8_out(letter, p);
*outix = index + ix;
return letter;
}
@@ -2532,6 +2556,8 @@ void InitNamedata(void)

void InitText2(void)
{
int param;

ungot_char = 0;
ungot_char2 = 0;

@@ -2539,7 +2565,7 @@ void InitText2(void)
n_param_stack = 1;
ssml_stack[0].tag_type = 0;

for (int param = 0; param < N_SPEECH_PARAM; param++)
for (param = 0; param < N_SPEECH_PARAM; param++)
speech_parameters[param] = param_stack[0].parameter[param]; // set all speech parameters to defaults

option_punctuation = speech_parameters[espeakPUNCTUATION];

+ 20
- 10
src/libespeak-ng/setlengths.c View File

@@ -140,7 +140,11 @@ static int speed3 = 118;
void SetSpeed(int control)
{
int x;
int s1;
int wpm;
int wpm2;
int wpm_value;
double sonic;

speed.loud_consonants = 0;
speed.min_sample_len = 450;
@@ -148,11 +152,11 @@ void SetSpeed(int control)
speed.lenmod2_factor = 100;
speed.min_pause = 5;

int wpm = embedded_value[EMBED_S];
wpm = embedded_value[EMBED_S];
if (control == 2)
wpm = embedded_value[EMBED_S2];

int wpm_value = wpm;
wpm_value = wpm;

if (voice->speed_percent > 0)
wpm = (wpm * voice->speed_percent)/100;
@@ -172,7 +176,7 @@ void SetSpeed(int control)
speed3 = (x * voice->speedf3)/256;
}
if (control & 2) {
double sonic = ((double)wpm2)/wpm;
sonic = ((double)wpm2)/wpm;
DoSonicSpeed((int)(sonic * 1024));
speed.pause_factor = 85;
speed.clause_pause_factor = 80;
@@ -225,7 +229,7 @@ void SetSpeed(int control)
speed.lenmod2_factor = 110 - (wpm - 250)/2;
}

int s1 = (x * voice->speedf1)/256;
s1 = (x * voice->speedf1)/256;

if (wpm >= 170)
speed.wav_factor = 110 + (150*s1)/128; // reduced speed adjustment, used for playing recorded sounds
@@ -269,13 +273,17 @@ void SetSpeed(int control)
void SetSpeed(int control)
{
// This is the earlier version of SetSpeed() before sonic speed-up was added
int x;
int s1;
int wpm;
int wpm2;

speed.loud_consonants = 0;
speed.min_sample_len = 450;
speed.lenmod_factor = 110; // controls the effect of FRFLAG_LEN_MOD reduce length change
speed.lenmod2_factor = 100;

int wpm = embedded_value[EMBED_S];
wpm = embedded_value[EMBED_S];
if (control == 2)
wpm = embedded_value[EMBED_S2];

@@ -287,10 +295,10 @@ void SetSpeed(int control)
if (wpm > 360)
speed.loud_consonants = (wpm - 360) / 8;

int wpm2 = wpm;
wpm2 = wpm;
if (wpm > 359) wpm2 = 359;
if (wpm < 80) wpm2 = 80;
int x = speed_lookup[wpm2-80];
x = speed_lookup[wpm2-80];

if (wpm >= 380)
x = 7;
@@ -321,7 +329,7 @@ void SetSpeed(int control)
speed.lenmod2_factor = 110 - (wpm - 250)/2;
}

int s1 = (x * voice->speedf1)/256;
s1 = (x * voice->speedf1)/256;

if (wpm >= 170)
speed.wav_factor = 110 + (150*s1)/128; // reduced speed adjustment, used for playing recorded sounds
@@ -365,10 +373,11 @@ void SetParameter(int parameter, int value, int relative)
// relative 0=absolute 1=relative

int new_value = value;
int default_value;

if (relative) {
if (parameter < 5) {
int default_value = param_defaults[parameter];
default_value = param_defaults[parameter];
new_value = default_value + (default_value * value)/100;
}
}
@@ -430,6 +439,7 @@ static void DoEmbedded2(int *embix)

void CalcLengths(Translator *tr)
{
int ix;
int ix2;
PHONEME_LIST *prev;
PHONEME_LIST *next;
@@ -458,7 +468,7 @@ void CalcLengths(Translator *tr)
unsigned char *pitch_env = NULL;
PHONEME_DATA phdata_tone;

for (int ix = 1; ix < n_phoneme_list; ix++) {
for (ix = 1; ix < n_phoneme_list; ix++) {
prev = &phoneme_list[ix-1];
p = &phoneme_list[ix];
stress = p->stresslevel & 0x7;

+ 13
- 6
src/libespeak-ng/speak_lib.c View File

@@ -264,6 +264,8 @@ static void init_path(const char *path)
{
#ifdef PLATFORM_WINDOWS
HKEY RegKey;
unsigned long size;
unsigned long var_type;
char *env;
unsigned char buf[sizeof(path_home)-13];

@@ -280,8 +282,8 @@ static void init_path(const char *path)

buf[0] = 0;
RegOpenKeyExA(HKEY_LOCAL_MACHINE, "Software\\Microsoft\\Speech\\Voices\\Tokens\\eSpeak", 0, KEY_READ, &RegKey);
unsigned long size = sizeof(buf);
unsigned long var_type = REG_SZ;
size = sizeof(buf);
var_type = REG_SZ;
RegQueryValueExA(RegKey, "path", 0, &var_type, buf, &size);

sprintf(path_home, "%s\\espeak-data", buf);
@@ -308,6 +310,7 @@ static void init_path(const char *path)

static int initialise(int control)
{
int param;
int result;
int srate = 22050; // default sample rate 22050 Hz

@@ -329,7 +332,7 @@ static int initialise(int control)
SynthesizeInit();
InitNamedata();

for (int param = 0; param < N_SPEECH_PARAM; param++)
for (param = 0; param < N_SPEECH_PARAM; param++)
param_stack[0].parameter[param] = param_defaults[param];

return 0;
@@ -444,6 +447,7 @@ void MarkerEvent(int type, unsigned int char_position, int value, int value2, un
{
// type: 1=word, 2=sentence, 3=named mark, 4=play audio, 5=end, 7=phoneme
espeak_EVENT *ep;
double time;

if ((event_list == NULL) || (event_list_ix >= (n_event_list-2)))
return;
@@ -455,7 +459,7 @@ void MarkerEvent(int type, unsigned int char_position, int value, int value2, un
ep->text_position = char_position & 0xffffff;
ep->length = char_position >> 24;

double time = ((double)(count_samples + mbrola_delay + (out_ptr - out_start)/2)*1000.0)/samplerate;
time = ((double)(count_samples + mbrola_delay + (out_ptr - out_start)/2)*1000.0)/samplerate;
ep->audio_position = (int)time;
ep->sample = (count_samples + mbrola_delay + (out_ptr - out_start)/2);

@@ -536,8 +540,9 @@ void sync_espeak_Key(const char *key)
{
// symbolic name, symbolicname_character - is there a system resource of symbolic names per language?
int letter;
int ix;

int ix = utf8_in(&letter, key);
ix = utf8_in(&letter, key);
if (key[ix] == 0) {
// a single character
sync_espeak_Char(letter);
@@ -595,6 +600,8 @@ ESPEAK_API void espeak_SetPhonemeCallback(int (*PhonemeCallback)(const char *))

ESPEAK_API int espeak_Initialize(espeak_AUDIO_OUTPUT output_type, int buf_length, const char *path, int options)
{
int param;

// It seems that the wctype functions don't work until the locale has been set
// to something other than the default "C". Then, not only Latin1 but also the
// other characters give the correct results with iswalpha() etc.
@@ -634,7 +641,7 @@ ESPEAK_API int espeak_Initialize(espeak_AUDIO_OUTPUT output_type, int buf_length

VoiceReset(0);

for (int param = 0; param < N_SPEECH_PARAM; param++)
for (param = 0; param < N_SPEECH_PARAM; param++)
param_stack[0].parameter[param] = saved_parameters[param] = param_defaults[param];

SetParameter(espeakRATE, 175, 0);

+ 8
- 5
src/libespeak-ng/spect.c View File

@@ -65,11 +65,11 @@ float polint(float xa[], float ya[], int n, float x)
{
// General polinomial interpolation routine, xa[1...n] ya[1...n]
int i, m, ns = 1;
float den, dift, ho, hp, w;
float den, dif, dift, ho, hp, w;
float y; // result
float c[9], d[9];

float dif = fabs(x-xa[1]);
dif = fabs(x-xa[1]);

for (i = 1; i <= n; i++) {
if ((dift = fabs(x-xa[i])) < dif) {
@@ -98,6 +98,7 @@ float polint(float xa[], float ya[], int n, float x)

static SpectFrame *SpectFrameCreate()
{
int ix;
SpectFrame *frame;

frame = malloc(sizeof(SpectFrame));
@@ -111,7 +112,7 @@ static SpectFrame *SpectFrameCreate()
frame->amp_adjust = 100;
frame->length_adjust = 0;

for (int ix = 0; ix < N_PEAKS; ix++) {
for (ix = 0; ix < N_PEAKS; ix++) {
frame->formants[ix].freq = 0;
frame->peaks[ix].pkfreq = default_freq[ix];
frame->peaks[ix].pkheight = 0;
@@ -199,6 +200,7 @@ double GetFrameRms(SpectFrame *frame, int seq_amplitude)
{
int h;
float total = 0;
int maxh;
int height;
int htab[400];
wavegen_peaks_t wpeaks[9];
@@ -212,7 +214,7 @@ double GetFrameRms(SpectFrame *frame, int seq_amplitude)
wpeaks[h].right = frame->peaks[h].pkright << 16;
}

int maxh = PeaksToHarmspect(wpeaks, 90<<16, htab, 0);
maxh = PeaksToHarmspect(wpeaks, 90<<16, htab, 0);
for (h = 1; h < maxh; h++)
total += ((htab[h] * htab[h]) >> 10);
frame->rms = sqrt(total) / 7.25;
@@ -243,8 +245,9 @@ SpectSeq *SpectSeqCreate()

void SpectSeqDestroy(SpectSeq *spect)
{
int ix;
if (spect->frames != NULL) {
for (int ix = 0; ix < spect->numframes; ix++) {
for (ix = 0; ix < spect->numframes; ix++) {
if (spect->frames[ix] != NULL)
SpectFrameDestroy(spect->frames[ix]);
}

+ 18
- 11
src/libespeak-ng/synth_mbrola.c View File

@@ -108,6 +108,8 @@ espeak_ERROR LoadMbrolaTable(const char *mbrola_voice, const char *phtrans, int
{
// Load a phoneme name translation table from espeak-data/mbrola

int size;
int ix;
int *pw;
FILE *f_in;
char path[sizeof(path_home)+15];
@@ -152,7 +154,7 @@ espeak_ERROR LoadMbrolaTable(const char *mbrola_voice, const char *phtrans, int

// read eSpeak's mbrola phoneme translation data, eg. en1_phtrans
sprintf(path, "%s/mbrola_ph/%s", path_home, phtrans);
int size = GetFileLength(path);
size = GetFileLength(path);
if ((f_in = fopen(path, "rb")) == NULL) {
close_MBR();
return EE_NOT_FOUND;
@@ -166,7 +168,7 @@ espeak_ERROR LoadMbrolaTable(const char *mbrola_voice, const char *phtrans, int

mbrola_control = Read4Bytes(f_in);
pw = (int *)mbrola_tab;
for (int ix = 4; ix < size; ix += 4)
for (ix = 4; ix < size; ix += 4)
*pw++ = Read4Bytes(f_in);
size = fread(mbrola_tab, 1, size, f_in);
fclose(f_in);
@@ -262,9 +264,11 @@ static int GetMbrName(PHONEME_LIST *plist, PHONEME_TAB *ph, PHONEME_TAB *ph_prev
static char *WritePitch(int env, int pitch1, int pitch2, int split, int final)
{
// final=1: only give the final pitch value.
int x;
int ix;
int pitch_base;
int pitch_range;
int p2;
int p1, p2, p_end;
unsigned char *pitch_env;
int max = -1;
int min = 999;
@@ -273,6 +277,7 @@ static char *WritePitch(int env, int pitch1, int pitch2, int split, int final)
int env100 = 80; // apply the pitch change only over this proportion of the mbrola phoneme(s)
int y2;
int y[4];
int env_split;
char buf[50];
static char output[50];

@@ -281,12 +286,12 @@ static char *WritePitch(int env, int pitch1, int pitch2, int split, int final)

SetPitch2(voice, pitch1, pitch2, &pitch_base, &pitch_range);

int env_split = (split * 128)/100;
env_split = (split * 128)/100;
if (env_split < 0)
env_split = 0-env_split;

// find max and min in the pitch envelope
for (int x = 0; x < 128; x++) {
for (x = 0; x < 128; x++) {
if (pitch_env[x] > max) {
max = pitch_env[x];
y_max = x;
@@ -307,8 +312,8 @@ static char *WritePitch(int env, int pitch1, int pitch2, int split, int final)
y[3] = y[2] + (127 - y[2])/2;

// set initial pitch
int p1 = ((pitch_env[0]*pitch_range)>>8) + pitch_base; // Hz << 12
int p_end = ((pitch_env[127]*pitch_range)>>8) + pitch_base;
p1 = ((pitch_env[0]*pitch_range)>>8) + pitch_base; // Hz << 12
p_end = ((pitch_env[127]*pitch_range)>>8) + pitch_base;

if (split >= 0) {
sprintf(buf, " 0 %d", p1/4096);
@@ -317,7 +322,7 @@ static char *WritePitch(int env, int pitch1, int pitch2, int split, int final)

// don't use intermediate pitch points for linear rise and fall
if (env > 1) {
for (int ix = 1; ix < 4; ix++) {
for (ix = 1; ix < 4; ix++) {
p2 = ((pitch_env[y[ix]]*pitch_range)>>8) + pitch_base;

if (split > 0)
@@ -574,20 +579,22 @@ int MbrolaFill(int length, int resume, int amplitude)
// Read audio data from Mbrola (length is in millisecs)

static int n_samples;
int req_samples, result;
int ix;
short value16;
int value;

if (!resume)
n_samples = samplerate * length / 1000;

int req_samples = (out_end - out_ptr)/2;
req_samples = (out_end - out_ptr)/2;
if (req_samples > n_samples)
req_samples = n_samples;
int result = read_MBR((short *)out_ptr, req_samples);
result = read_MBR((short *)out_ptr, req_samples);
if (result <= 0)
return 0;

for (int ix = 0; ix < result; ix++) {
for (ix = 0; ix < result; ix++) {
value16 = out_ptr[0] + (out_ptr[1] << 8);
value = value16 * amplitude;
value = value / 40; // adjust this constant to give a suitable amplitude for mbrola voices

+ 39
- 21
src/libespeak-ng/synthdata.c View File

@@ -75,10 +75,11 @@ static char *ReadPhFile(void *ptr, const char *fname, int *size)
{
FILE *f_in;
char *p;
unsigned int length;
char buf[sizeof(path_home)+40];

sprintf(buf, "%s%c%s", path_home, PATHSEP, fname);
unsigned int length = GetFileLength(buf);
length = GetFileLength(buf);

if ((f_in = fopen(buf, "rb")) == NULL) {
fprintf(stderr, "Can't read data file: '%s'\n", buf);
@@ -108,8 +109,10 @@ int LoadPhData(int *srate)
{
int ix;
int n_phonemes;
int version;
int result = 1;
int length;
int rate;
unsigned char *p;
int *pw;

@@ -125,9 +128,8 @@ int LoadPhData(int *srate)
n_tunes = length / sizeof(TUNE);

// read the version number and sample rate from the first 8 bytes of phondata
int version = 0; // bytes 0-3, version number
int rate = 0; // bytes 4-7, sample rate

version = 0; // bytes 0-3, version number
rate = 0; // bytes 4-7, sample rate
for (ix = 0; ix < 4; ix++) {
version += (wavefile_data[ix] << (ix*8));
rate += (wavefile_data[ix+4] << (ix*8));
@@ -176,7 +178,9 @@ void FreePhData(void)

int PhonemeCode(unsigned int mnem)
{
for (int ix = 0; ix < n_phoneme_tab; ix++) {
int ix;

for (ix = 0; ix < n_phoneme_tab; ix++) {
if (phoneme_tab[ix] == NULL)
continue;
if (phoneme_tab[ix]->mnemonic == mnem)
@@ -187,11 +191,13 @@ int PhonemeCode(unsigned int mnem)

int LookupPhonemeString(const char *string)
{
int ix;
unsigned char c;
unsigned int mnem;

// Pack up to 4 characters into a word
unsigned int mnem = 0;
for (int ix = 0; ix < 4; ix++) {
mnem = 0;
for (ix = 0; ix < 4; ix++) {
if (string[ix] == 0) break;
c = string[ix];
mnem |= (c << (ix*8));
@@ -203,7 +209,9 @@ int LookupPhonemeString(const char *string)
frameref_t *LookupSpect(PHONEME_TAB *this_ph, int which, FMT_PARAMS *fmt_params, int *n_frames, PHONEME_LIST *plist)
{
int ix;
int nf;
int nf1;
int seq_break;
frameref_t *frames;
int length1;
int length_std;
@@ -215,13 +223,13 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph, int which, FMT_PARAMS *fmt_params,

seq = (SPECT_SEQ *)(&phondata_ptr[fmt_params->fmt_addr]);
seqk = (SPECT_SEQK *)seq;
int nf = seq->n_frames;
nf = seq->n_frames;

if (nf >= N_SEQ_FRAMES)
nf = N_SEQ_FRAMES - 1;

seq_len_adjust = fmt_params->fmt2_lenadj + fmt_params->fmt_length;
int seq_break = 0;
seq_break = 0;

for (ix = 0; ix < nf; ix++) {
if (seq->frame[0].frflags & FRFLAG_KLATT)
@@ -333,6 +341,7 @@ unsigned char *GetEnvelope(int index)

static void SetUpPhonemeTable(int number, int recursing)
{
int ix;
int includes;
int ph_code;
PHONEME_TAB *phtab;
@@ -347,7 +356,7 @@ static void SetUpPhonemeTable(int number, int recursing)

// now add the phonemes from this table
phtab = phoneme_tab_list[number].phoneme_tab_ptr;
for (int ix = 0; ix < phoneme_tab_list[number].n_phonemes; ix++) {
for (ix = 0; ix < phoneme_tab_list[number].n_phonemes; ix++) {
ph_code = phtab[ix].code;
phoneme_tab[ph_code] = &phtab[ix];
if (ph_code > n_phoneme_tab)
@@ -455,6 +464,7 @@ static bool StressCondition(Translator *tr, PHONEME_LIST *plist, int condition,
// condition:
// 0 if diminished, 1 if unstressed, 2 if not stressed, 3 if stressed, 4 if max stress

int stress_level;
PHONEME_LIST *pl;
static int condition_level[4] = { 1, 2, 4, 15 };

@@ -468,7 +478,7 @@ static bool StressCondition(Translator *tr, PHONEME_LIST *plist, int condition,
return false; // no stress elevel for this consonant
}

int stress_level = pl->stresslevel & 0xf;
stress_level = pl->stresslevel & 0xf;

if (tr != NULL) {
if ((control & 1) && (plist->synthflags & SFLAG_DICTIONARY) && ((tr->langopts.param[LOPT_REDUCE] & 1) == 0)) {
@@ -515,6 +525,9 @@ static bool InterpretCondition(Translator *tr, int control, PHONEME_LIST *plist,
{
int which;
int ix;
unsigned int data;
int instn;
int instn2;
int count;
int check_endtype = 0;
PHONEME_TAB *ph;
@@ -528,9 +541,9 @@ static bool InterpretCondition(Translator *tr, int control, PHONEME_LIST *plist,

// bits 8-10 = 7, other conditions

int instn = (*p_prog) & 0xfff;
unsigned int data = instn & 0xff;
int instn2 = instn >> 8;
instn = (*p_prog) & 0xfff;
data = instn & 0xff;
instn2 = instn >> 8;

if (instn2 < 14) {
plist_this = plist;
@@ -736,6 +749,7 @@ static void SwitchOnVowelType(PHONEME_LIST *plist, PHONEME_DATA *phdata, USHORT
{
USHORT *prog;
int voweltype;
signed char x;

if (instn_type == 2) {
phdata->pd_control |= pd_FORNEXTPH;
@@ -747,7 +761,7 @@ static void SwitchOnVowelType(PHONEME_LIST *plist, PHONEME_DATA *phdata, USHORT
if ((voweltype >= 0) && (voweltype < 6)) {
prog = *p_prog + voweltype*2;
phdata->sound_addr[instn_type] = (((prog[1] & 0xf) << 16) + prog[2]) * 4;
signed char x = (prog[1] >> 4) & 0xff;
x = (prog[1] >> 4) & 0xff;
phdata->sound_param[instn_type] = x; // sign extend
}

@@ -756,13 +770,15 @@ static void SwitchOnVowelType(PHONEME_LIST *plist, PHONEME_DATA *phdata, USHORT

int NumInstnWords(USHORT *prog)
{
int instn;
int instn2;
int instn_type;
int n;
int type2;
static const char n_words[16] = { 0, 1, 0, 0, 1, 1, 0, 1, 1, 2, 4, 0, 0, 0, 0, 0 };

int instn2;
int type2;
int instn = *prog;
int instn_type = instn >> 12;
instn = *prog;
instn_type = instn >> 12;
if ((n = n_words[instn_type]) > 0)
return n;

@@ -812,6 +828,7 @@ void InterpretPhoneme(Translator *tr, int control, PHONEME_LIST *plist, PHONEME_
bool truth;
bool truth2;
int data;
int end_flag;
int ix;
signed char param_sc;

@@ -833,7 +850,7 @@ void InterpretPhoneme(Translator *tr, int control, PHONEME_LIST *plist, PHONEME_
if (ph->program == 0)
return;

int end_flag = 0;
end_flag = 0;

for (prog = &phoneme_index[ph->program]; end_flag != 1; prog++) {
instn = *prog;
@@ -1042,10 +1059,11 @@ void InterpretPhoneme(Translator *tr, int control, PHONEME_LIST *plist, PHONEME_
void InterpretPhoneme2(int phcode, PHONEME_DATA *phdata)
{
// Examine the program of a single isolated phoneme
int ix;
PHONEME_LIST plist[4];
memset(plist, 0, sizeof(plist));

for (int ix = 0; ix < 4; ix++) {
for (ix = 0; ix < 4; ix++) {
plist[ix].phcode = phonPAUSE;
plist[ix].ph = phoneme_tab[phonPAUSE];
}

+ 62
- 30
src/libespeak-ng/synthesize.c View File

@@ -77,9 +77,10 @@ static PHONEME_LIST next_pause;
const char *WordToString(unsigned int word)
{
// Convert a phoneme mnemonic word into a string
int ix;
static char buf[5];

for (int ix = 0; ix < 4; ix++)
for (ix = 0; ix < 4; ix++)
buf[ix] = word >> (ix*8);
buf[4] = 0;
return buf;
@@ -189,6 +190,7 @@ static void DoPause(int length, int control)
// control = 1, less shortening at fast speeds

unsigned int len;
int srate2;

if (length == 0)
len = 0;
@@ -198,7 +200,7 @@ static void DoPause(int length, int control)
if (len < 90000)
len = (len * samplerate) / 1000; // convert from mS to number of samples
else {
int srate2 = samplerate / 25; // avoid overflow
srate2 = samplerate / 25; // avoid overflow
len = (len * srate2) / 40;
}
}
@@ -220,20 +222,25 @@ extern int seq_len_adjust; // temporary fix to advance the start point for playi

static int DoSample2(int index, int which, int std_length, int control, int length_mod, int amp)
{
int length;
int wav_length;
int wav_scale;
int min_length;
int x;
int len4;
intptr_t *q;
unsigned char *p;

index = index & 0x7fffff;
p = &wavefile_data[index];
int wav_scale = p[2];
int wav_length = (p[1] * 256);
wav_scale = p[2];
wav_length = (p[1] * 256);
wav_length += p[0]; // length in bytes

if (wav_length == 0)
return 0;

int min_length = speed.min_sample_len;
min_length = speed.min_sample_len;

if (wav_scale == 0)
min_length *= 2; // 16 bit samples
@@ -254,7 +261,7 @@ static int DoSample2(int index, int which, int std_length, int control, int leng
if (length_mod > 0)
std_length = (std_length * length_mod)/256;

int length = (std_length * speed.wav_factor)/256;
length = (std_length * speed.wav_factor)/256;

if (control & pd_DONTLENGTHEN) {
// this option is used for Stops, with short noise bursts.
@@ -278,7 +285,7 @@ static int DoSample2(int index, int which, int std_length, int control, int leng
if (amp < 0)
return length;

int len4 = wav_length / 4;
len4 = wav_length / 4;

index += 4;

@@ -389,6 +396,10 @@ static void set_frame_rms(frame_t *fr, int new_rms)
// Each frame includes its RMS amplitude value, so to set a new
// RMS just adjust the formant amplitudes by the appropriate ratio

int x;
int h;
int ix;

static const short sqrt_tab[200] = {
0, 64, 90, 110, 128, 143, 156, 169, 181, 192, 202, 212, 221, 230, 239, 247,
256, 263, 271, 278, 286, 293, 300, 306, 313, 320, 326, 332, 338, 344, 350, 356,
@@ -412,13 +423,12 @@ static void set_frame_rms(frame_t *fr, int new_rms)
}

if (fr->rms == 0) return; // check for divide by zero
int x = (new_rms * 64)/fr->rms;
x = (new_rms * 64)/fr->rms;
if (x >= 200) x = 199;

x = sqrt_tab[x]; // sqrt(new_rms/fr->rms)*0x200;

int h;
for (int ix = 0; ix < 8; ix++) {
for (ix = 0; ix < 8; ix++) {
h = fr->fheight[ix] * x;
fr->fheight[ix] = h/0x200;
}
@@ -427,12 +437,13 @@ static void set_frame_rms(frame_t *fr, int new_rms)
static void formants_reduce_hf(frame_t *fr, int level)
{
// change height of peaks 2 to 8, percentage
int ix;
int x;

if (voice->klattv[0])
return;

int x;
for (int ix = 2; ix < 8; ix++) {
for (ix = 2; ix < 8; ix++) {
x = fr->fheight[ix] * level;
fr->fheight[ix] = x/100;
}
@@ -471,9 +482,11 @@ static frame_t *DuplicateLastFrame(frameref_t *seq, int n_frames, int length)

static void AdjustFormants(frame_t *fr, int target, int min, int max, int f1_adj, int f3_adj, int hf_reduce, int flags)
{
int x;

target = (target * voice->formant_factor)/256;

int x = (target - fr->ffreq[2]) / 2;
x = (target - fr->ffreq[2]) / 2;
if (x > max) x = max;
if (x < min) x = min;
fr->ffreq[2] += x;
@@ -523,6 +536,21 @@ static int VowelCloseness(frame_t *fr)

int FormantTransition2(frameref_t *seq, int *n_frames, unsigned int data1, unsigned int data2, PHONEME_TAB *other_ph, int which)
{
int ix;
int formant;
int next_rms;

int len;
int rms;
int f1;
int f2;
int f2_min;
int f2_max;
int f3_adj;
int f3_amp;
int flags;
int vcolour;

#define N_VCOLOUR 2
// percentage change for each formant in 256ths
static short vcolouring[N_VCOLOUR][5] = {
@@ -535,17 +563,17 @@ int FormantTransition2(frameref_t *seq, int *n_frames, unsigned int data1, unsig
if (*n_frames < 2)
return 0;

int len = (data1 & 0x3f) * 2;
int rms = (data1 >> 6) & 0x3f;
int flags = (data1 >> 12);
len = (data1 & 0x3f) * 2;
rms = (data1 >> 6) & 0x3f;
flags = (data1 >> 12);

int f2 = (data2 & 0x3f) * 50;
int f2_min = (((data2 >> 6) & 0x1f) - 15) * 50;
int f2_max = (((data2 >> 11) & 0x1f) - 15) * 50;
int f3_adj = (((data2 >> 16) & 0x1f) - 15) * 50;
int f3_amp = ((data2 >> 21) & 0x1f) * 8;
int f1 = ((data2 >> 26) & 0x7);
int vcolour = (data2 >> 29);
f2 = (data2 & 0x3f) * 50;
f2_min = (((data2 >> 6) & 0x1f) - 15) * 50;
f2_max = (((data2 >> 11) & 0x1f) - 15) * 50;
f3_adj = (((data2 >> 16) & 0x1f) - 15) * 50;
f3_amp = ((data2 >> 21) & 0x1f) * 8;
f1 = ((data2 >> 26) & 0x7);
vcolour = (data2 >> 29);

if ((other_ph != NULL) && (other_ph->mnemonic == '?'))
flags |= 8;
@@ -560,7 +588,7 @@ int FormantTransition2(frameref_t *seq, int *n_frames, unsigned int data1, unsig
seq[0].frflags |= FRFLAG_LEN_MOD2; // reduce length modification
fr->frflags |= FRFLAG_LEN_MOD2;

int next_rms = seq[1].frame->rms;
next_rms = seq[1].frame->rms;

if (voice->klattv[0])
fr->klattp[KLATT_AV] = seq[1].frame->klattp[KLATT_AV] - 4;
@@ -604,11 +632,11 @@ int FormantTransition2(frameref_t *seq, int *n_frames, unsigned int data1, unsig
set_frame_rms(fr, rms);

if ((vcolour > 0) && (vcolour <= N_VCOLOUR)) {
for (int ix = 0; ix < *n_frames; ix++) {
for (ix = 0; ix < *n_frames; ix++) {
fr = CopyFrame(seq[ix].frame, 0);
seq[ix].frame = fr;

for (int formant = 1; formant <= 5; formant++) {
for (formant = 1; formant <= 5; formant++) {
int x;
x = fr->ffreq[formant] * vcolouring[vcolour-1][formant-1];
fr->ffreq[formant] = x / 256;
@@ -642,6 +670,7 @@ static void SmoothSpect(void)
frame_t *frame2;
frame_t *frame1;
frame_t *frame_centre;
int ix;
int len;
int pk;
int modified;
@@ -660,7 +689,7 @@ static void SmoothSpect(void)
frame_centre = (frame_t *)q[2];

// backwards
int ix = syllable_centre -1;
ix = syllable_centre -1;
frame = frame2 = frame_centre;
for (;;) {
if (ix < 0) ix = N_WCMDQ-1;
@@ -826,6 +855,9 @@ int DoSpect2(PHONEME_TAB *this_ph, int which, FMT_PARAMS *fmt_params, PHONEME_L
int len;
int frame_length;
int length_factor;
int length_mod;
int length_sum;
int length_min;
int total_len = 0;
static int wave_flag = 0;
int wcmd_spect = WCMD_SPECT;
@@ -834,10 +866,10 @@ int DoSpect2(PHONEME_TAB *this_ph, int which, FMT_PARAMS *fmt_params, PHONEME_L
if (fmt_params->fmt_addr == 0)
return 0;

int length_mod = plist->length;
length_mod = plist->length;
if (length_mod == 0) length_mod = 256;

int length_min = (samplerate/70); // greater than one cycle at low pitch (Hz)
length_min = (samplerate/70); // greater than one cycle at low pitch (Hz)
if (which == 2) {
if ((translator->langopts.param[LOPT_LONG_VOWEL_THRESHOLD] > 0) && ((this_ph->std_length >= translator->langopts.param[LOPT_LONG_VOWEL_THRESHOLD]) || (plist->synthflags & SFLAG_LENGTHEN) || (this_ph->phflags & phLONG)))
length_min *= 2; // ensure long vowels are longer
@@ -915,7 +947,7 @@ int DoSpect2(PHONEME_TAB *this_ph, int which, FMT_PARAMS *fmt_params, PHONEME_L
syllable_centre = wcmdq_tail;
}

int length_sum = 0;
length_sum = 0;
for (frameix = 1; frameix < n_frames; frameix++) {
length_factor = length_mod;
if (frames[frameix-1].frflags & FRFLAG_LEN_MOD) // reduce effect of length mod

+ 14
- 6
src/libespeak-ng/tr_languages.c View File

@@ -143,25 +143,32 @@ static void SetLetterVowel(Translator *tr, int c)
static void ResetLetterBits(Translator *tr, int groups)
{
// Clear all the specified groups
unsigned int mask = ~groups;
unsigned int ix;
unsigned int mask;

for (int ix = 0; ix < sizeof(tr->letter_bits); ix++)
mask = ~groups;

for (ix = 0; ix < sizeof(tr->letter_bits); ix++)
tr->letter_bits[ix] &= mask;
}

static void SetLetterBits(Translator *tr, int group, const char *string)
{
int bits;
unsigned char c;

int bits = (1L << group);
bits = (1L << group);
while ((c = *string++) != 0)
tr->letter_bits[c] |= bits;
}

static void SetLetterBitsRange(Translator *tr, int group, int first, int last)
{
int bits = (1L << group);
for (int ix = first; ix <= last; ix++)
int bits;
int ix;

bits = (1L << group);
for (ix = first; ix <= last; ix++)
tr->letter_bits[ix] |= bits;
}

@@ -187,6 +194,7 @@ const char string_ordinal[] = { 0xc2, 0xba, 0 }; // masculine ordinal character,
static Translator *NewTranslator(void)
{
Translator *tr;
int ix;
static const unsigned char stress_amps2[] = { 18, 18, 20, 20, 20, 22, 22, 20 };
static const short stress_lengths2[8] = { 182, 140, 220, 220, 220, 240, 260, 280 };
static const wchar_t empty_wstring[1] = { 0 };
@@ -254,7 +262,7 @@ static Translator *NewTranslator(void)
tr->punct_within_word = punct_in_word;
tr->chars_ignore = chars_ignore_default;

for (int ix = 0; ix < 8; ix++) {
for (ix = 0; ix < 8; ix++) {
tr->stress_amps[ix] = stress_amps2[ix];
tr->stress_amps_r[ix] = stress_amps2[ix] - 1;
tr->stress_lengths[ix] = stress_lengths2[ix];

+ 34
- 16
src/libespeak-ng/translate.c View File

@@ -458,8 +458,9 @@ void DeleteTranslator(Translator *tr)
int lookupwchar(const unsigned short *list, int c)
{
// Is the character c in the list ?
int ix;

for (int ix = 0; list[ix] != 0; ix++) {
for (ix = 0; list[ix] != 0; ix++) {
if (list[ix] == c)
return ix+1;
}
@@ -471,7 +472,9 @@ int lookupwchar2(const unsigned short *list, int c)
// Replace character c by another character.
// Returns 0 = not found, 1 = delete character

for (int ix = 0; list[ix] != 0; ix += 2) {
int ix;

for (ix = 0; list[ix] != 0; ix += 2) {
if (list[ix] == c)
return list[ix+1];
}
@@ -491,6 +494,8 @@ int utf8_out(unsigned int c, char *buf)
// returns the number of bytes written

int n_bytes;
int j;
int shift;
static char unsigned code[4] = { 0, 0xc0, 0xe0, 0xf0 };

if (c < 0x80) {
@@ -508,9 +513,9 @@ int utf8_out(unsigned int c, char *buf)
else
n_bytes = 3;

int shift = 6*n_bytes;
shift = 6*n_bytes;
buf[0] = code[n_bytes] | (c >> shift);
for (int j = 0; j < n_bytes; j++) {
for (j = 0; j < n_bytes; j++) {
shift -= 6;
buf[j+1] = 0x80 + ((c >> shift) & 0x3f);
}
@@ -538,6 +543,8 @@ int utf8_in2(int *c, const char *buf, int backwards)
// backwards: set if we are moving backwards through the UTF8 string

int c1;
int n_bytes;
int ix;
static const unsigned char mask[4] = { 0xff, 0x1f, 0x0f, 0x07 };

// find the start of the next/previous character
@@ -549,7 +556,7 @@ int utf8_in2(int *c, const char *buf, int backwards)
buf++;
}

int n_bytes = 0;
n_bytes = 0;

if ((c1 = *buf++) & 0x80) {
if ((c1 & 0xe0) == 0xc0)
@@ -560,7 +567,7 @@ int utf8_in2(int *c, const char *buf, int backwards)
n_bytes = 3;

c1 &= mask[n_bytes];
for (int ix = 0; ix < n_bytes; ix++)
for (ix = 0; ix < n_bytes; ix++)
c1 = (c1 << 6) + (*buf++ & 0x3f);
}
*c = c1;
@@ -626,13 +633,14 @@ static int CheckDottedAbbrev(char *word1, WORD_TAB *wtab)
int count = 0;
int nbytes;
int ok;
int ix;
char *word;
char *wbuf;
char word_buf[80];

word = word1;
wbuf = word_buf;
int ix = 0;
ix = 0;

for (;;) {
ok = 0;
@@ -691,6 +699,7 @@ int ChangeEquivalentPhonemes(Translator *tr, int lang2, char *phonemes)
char *eqlist;
char *p_out;
char *p_in;
int remove_stress = 0;
char phonbuf[N_WORD_PHONEMES];

// has a phoneme equivalence table been specified for this language pair?
@@ -709,7 +718,7 @@ int ChangeEquivalentPhonemes(Translator *tr, int lang2, char *phonemes)
len = (pb[2] << 8) + pb[3]; // size of this table in words
pb += (len * 4);
}
int remove_stress = pb[1];
remove_stress = pb[1];

if (option_phonemes & espeakPHONEMES_TRACE) {
DecodePhonemes(phonemes, phonbuf);
@@ -760,6 +769,7 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
// word1 is terminated by space (0x20) character

char *word1;
int word_length;
int ix;
char *p;
int pfix;
@@ -779,6 +789,7 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
char end_phonemes2[N_WORD_PHONEMES];
char word_copy[N_WORD_BYTES];
char word_copy2[N_WORD_BYTES];
int word_copy_length;
char prefix_chars[0x3f + 2];
int found = 0;
int end_flags;
@@ -792,6 +803,8 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
int spell_word;
int stress_bits;
int emphasize_allcaps = 0;
int wflags;
int wmark;
int was_unpronouncable = 0;
int loopcount;
int add_suffix_phonemes = 0;
@@ -807,8 +820,8 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
memset(wtab_null, 0, sizeof(wtab_null));
wtab = wtab_null;
}
int wflags = wtab->flags;
int wmark = wtab->wmark;
wflags = wtab->flags;
wmark = wtab->wmark;

dictionary_flags[0] = 0;
dictionary_flags[1] = 0;
@@ -833,13 +846,13 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
wordx = word1;

utf8_in(&first_char, wordx);
int word_length = 0;
word_length = 0;
while ((*wordx != 0) && (*wordx != ' ')) {
wordx += utf8_in(&last_char, wordx);
word_length++;
}

int word_copy_length = wordx - word_start;
word_copy_length = wordx - word_start;
if (word_copy_length >= N_WORD_BYTES)
word_copy_length = N_WORD_BYTES-1;
memcpy(word_copy2, word_start, word_copy_length);
@@ -987,9 +1000,13 @@ int TranslateWord(Translator *tr, char *word_start, int next_pause, WORD_TAB *wt
// word's pronunciation is not given in the dictionary list, although
// dictionary_flags may have ben set there

int posn = 0;
int non_initial = 0;
int length = 999;
int posn;
int non_initial;
int length;

posn = 0;
non_initial = 0;
length = 999;
wordx = word1;

while (((length < 3) && (length > 0)) || (word_length > 1 && Unpronouncable(tr, wordx, posn))) {
@@ -2149,12 +2166,13 @@ static const char *UCase_ga[] = { "bp", "bhf", "dt", "gc", "hA", "mb", "nd", "ng

int UpperCaseInWord(Translator *tr, char *word, int c)
{
int ix;
int len;
const char *p;

if (tr->translator_name == L('g', 'a')) {
// Irish
for (int ix = 0;; ix++) {
for (ix = 0;; ix++) {
if ((p = UCase_ga[ix]) == NULL)
break;


+ 29
- 15
src/libespeak-ng/voices.c View File

@@ -256,8 +256,9 @@ static void SetToneAdjust(voice_t *voice, int *tone_pts)
void ReadTonePoints(char *string, int *tone_pts)
{
// tone_pts[] is int[12]
int ix;

for (int ix = 0; ix < 12; ix++)
for (ix = 0; ix < 12; ix++)
tone_pts[ix] = -1;

sscanf(string, "%d %d %d %d %d %d %d %d %d %d",
@@ -283,7 +284,9 @@ static espeak_VOICE *ReadVoiceFile(FILE *f_in, const char *fname, const char *le
char *p;
espeak_VOICE *voice_data;
int priority;
int age;
int n_variants = 4; // default, number of variants of this voice before using another voice
int gender;

#ifdef PLATFORM_WINDOWS
char fname_buf[sizeof(path_home)+15];
@@ -300,7 +303,7 @@ static espeak_VOICE *ReadVoiceFile(FILE *f_in, const char *fname, const char *le

vname[0] = 0;
vgender[0] = 0;
int age = 0;
age = 0;

while (fgets_strip(linebuf, sizeof(linebuf), f_in) != NULL) {
if (memcmp(linebuf, "name", 4) == 0) {
@@ -328,7 +331,7 @@ static espeak_VOICE *ReadVoiceFile(FILE *f_in, const char *fname, const char *le
}
languages[langix++] = 0;

int gender = LookupMnem(genders, vgender);
gender = LookupMnem(genders, vgender);

if (n_languages == 0)
return NULL; // no language lines in the voice file
@@ -361,6 +364,7 @@ void VoiceReset(int tone_only)
{
// Set voice to the default values

int pk;
static unsigned char default_heights[N_PEAKS] = { 130, 128, 120, 116, 100, 100, 128, 128, 128 }; // changed for v.1.47
static unsigned char default_widths[N_PEAKS] = { 140, 128, 128, 160, 171, 171, 128, 128, 128 };

@@ -391,7 +395,7 @@ void VoiceReset(int tone_only)
voice->roughness = 2;

InitBreath();
for (int pk = 0; pk < N_PEAKS; pk++) {
for (pk = 0; pk < N_PEAKS; pk++) {
voice->freq[pk] = 256;
voice->height[pk] = default_heights[pk]*2;
voice->width[pk] = default_widths[pk]*2;
@@ -423,13 +427,14 @@ void VoiceReset(int tone_only)
static void VoiceFormant(char *p)
{
// Set parameters for a formant
int ix;
int formant;
int freq = 100;
int height = 100;
int width = 100;
int freqadd = 0;

int ix = sscanf(p, "%d %d %d %d %d", &formant, &freq, &height, &width, &freqadd);
ix = sscanf(p, "%d %d %d %d %d", &formant, &freq, &height, &width, &freqadd);
if (ix < 2)
return;

@@ -447,13 +452,14 @@ static void VoiceFormant(char *p)

static void PhonemeReplacement(int type, char *p)
{
int n;
int phon;
int flags = 0;
char phon_string1[12];
char phon_string2[12];

strcpy(phon_string2, "NULL");
int n = sscanf(p, "%d %s %s", &flags, phon_string1, phon_string2);
n = sscanf(p, "%d %s %s", &flags, phon_string1, phon_string2);
if ((n < 2) || (n_replace_phonemes >= N_REPLACE_PHONEMES))
return;

@@ -476,10 +482,11 @@ static int Read8Numbers(char *data_in, int *data)
static unsigned int StringToWord2(const char *string)
{
// Convert a language name string to a word such as L('e','n')
int ix;
int c;
unsigned int value = 0;

for (int ix = 0; (ix < 4) && ((c = string[ix]) != 0); ix++)
for (ix = 0; (ix < 4) && ((c = string[ix]) != 0); ix++)
value = (value << 8) | (c & 0xff);
return value;
}
@@ -1175,6 +1182,7 @@ static int ScoreVoice(espeak_VOICE *voice_spec, const char *spec_language, int s
static int SetVoiceScores(espeak_VOICE *voice_select, espeak_VOICE **voices, int control)
{
// control: bit0=1 include mbrola voices
int ix;
int score;
int nv; // number of candidates
int n_parts = 0;
@@ -1187,7 +1195,7 @@ static int SetVoiceScores(espeak_VOICE *voice_select, espeak_VOICE **voices, int
if ((voice_select->languages != NULL) && (voice_select->languages[0] != 0)) {
n_parts = 1;
lang_len = strlen(voice_select->languages);
for (int ix = 0; (ix <= lang_len) && ((unsigned)ix < sizeof(language)); ix++) {
for (ix = 0; (ix <= lang_len) && ((unsigned)ix < sizeof(language)); ix++) {
if ((language[ix] = tolower(voice_select->languages[ix])) == '-')
n_parts++;
}
@@ -1210,7 +1218,7 @@ static int SetVoiceScores(espeak_VOICE *voice_select, espeak_VOICE **voices, int

// select those voices which match the specified language
nv = 0;
for (int ix = 0; ix < n_voices_list; ix++) {
for (ix = 0; ix < n_voices_list; ix++) {
vp = voices_list[ix];

if (((control & 1) == 0) && (memcmp(vp->identifier, "mb/", 3) == 0))
@@ -1234,11 +1242,13 @@ static int SetVoiceScores(espeak_VOICE *voice_select, espeak_VOICE **voices, int

espeak_VOICE *SelectVoiceByName(espeak_VOICE **voices, const char *name2)
{
int ix;
int match_fname = -1;
int match_fname2 = -1;
int match_name = -1;
const char *id; // this is the filename within espeak-data/voices
char *variant_name;
int last_part_len;
char last_part[41];
char name[40];

@@ -1255,9 +1265,9 @@ espeak_VOICE *SelectVoiceByName(espeak_VOICE **voices, const char *name2)
}

sprintf(last_part, "%c%s", PATHSEP, name);
int last_part_len = strlen(last_part);
last_part_len = strlen(last_part);

for (int ix = 0; voices[ix] != NULL; ix++) {
for (ix = 0; voices[ix] != NULL; ix++) {
if (strcmp(name, voices[ix]->name) == 0) {
match_name = ix; // found matching voice name
break;
@@ -1497,6 +1507,7 @@ static void GetVoices(const char *path)
espeak_ERROR SetVoiceByName(const char *name)
{
espeak_VOICE *v;
int ix;
espeak_VOICE voice_selector;
char *variant_name;
static char buf[60];
@@ -1505,7 +1516,7 @@ espeak_ERROR SetVoiceByName(const char *name)

variant_name = ExtractVoiceVariantName(buf, 0, 1);

for (int ix = 0;; ix++) {
for (ix = 0;; ix++) {
// convert voice name to lower case (ascii)
if ((buf[ix] = tolower(buf[ix])) == 0)
break;
@@ -1562,7 +1573,8 @@ espeak_ERROR SetVoiceByProperties(espeak_VOICE *voice_selector)

void FreeVoiceList()
{
for (int ix = 0; ix < n_voices_list; ix++) {
int ix;
for (ix = 0; ix < n_voices_list; ix++) {
if (voices_list[ix] != NULL) {
free(voices_list[ix]);
voices_list[ix] = NULL;
@@ -1577,6 +1589,8 @@ ESPEAK_API const espeak_VOICE **espeak_ListVoices(espeak_VOICE *voice_spec)
{
char path_voices[sizeof(path_home)+12];

int ix;
int j;
espeak_VOICE *v;
static espeak_VOICE **voices = NULL;

@@ -1599,8 +1613,8 @@ ESPEAK_API const espeak_VOICE **espeak_ListVoices(espeak_VOICE *voice_spec)
SetVoiceScores(voice_spec, voices, 1);
} else {
// list all: omit variant voices and mbrola voices and test voices
int j = 0;
for (int ix = 0; (v = voices_list[ix]) != NULL; ix++) {
j = 0;
for (ix = 0; (v = voices_list[ix]) != NULL; ix++) {
if ((v->languages[0] != 0) && (strcmp(&v->languages[1], "variant") != 0)
&& (memcmp(v->identifier, "mb/", 3) != 0) && (memcmp(v->identifier, "test/", 5) != 0))
voices[j++] = v;

+ 4
- 2
src/libespeak-ng/wave_sada.c View File

@@ -170,6 +170,7 @@ size_t wave_write(void *theHandler,
char *theMono16BitsWaveBuffer,
size_t theSize)
{
size_t num;
if (my_callback_is_output_enabled && (0 == my_callback_is_output_enabled()))
return 0;

@@ -188,7 +189,7 @@ size_t wave_write(void *theHandler,
}
#endif

size_t num = write((int)theHandler, theMono16BitsWaveBuffer, theSize);
num = write((int)theHandler, theMono16BitsWaveBuffer, theSize);

// Keep track of the total number of samples sent -- we use this in
// wave_get_read_position and also use it to help calculate the
@@ -228,12 +229,13 @@ size_t wave_write(void *theHandler,
//
int wave_close(void *theHandler)
{
int ret;
audio_info_t ainfo;
int audio_fd = (int)theHandler;
if (!audio_fd)
audio_fd = sun_audio_fd;
// [[[WDW: maybe do a pause/resume ioctl???]]]
int ret = ioctl(audio_fd, I_FLUSH, FLUSHRW);
ret = ioctl(audio_fd, I_FLUSH, FLUSHRW);
ioctl(audio_fd, AUDIO_GETINFO, &ainfo);

// Calculate the number of samples that won't get

+ 72
- 34
src/libespeak-ng/wavegen.c View File

@@ -280,7 +280,8 @@ void WcmdqStop()

int WcmdqFree()
{
int i = wcmdq_head - wcmdq_tail;
int i;
i = wcmdq_head - wcmdq_tail;
if (i <= 0) i += N_WCMDQ;
return i;
}
@@ -365,14 +366,17 @@ static int WaveCallback(const void *inputBuffer, void *outputBuffer,
PaStreamCallbackFlags flags, void *userData)
#endif
{
int ix;
int result;
unsigned char *p;
unsigned char *out_buf;
unsigned char *out_end2;
int pa_size;

int pa_size = framesPerBuffer*2;
pa_size = framesPerBuffer*2;

// make a buffer 3x size of the portaudio output
int ix = pa_size*3;
ix = pa_size*3;
if (ix > outbuffer_size) {
outbuffer = (unsigned char *)realloc(outbuffer, ix);
if (outbuffer == NULL) {
@@ -389,7 +393,7 @@ static int WaveCallback(const void *inputBuffer, void *outputBuffer,

event_list_ix = 0;

int result = WavegenFill(1);
result = WavegenFill(1);

// copy from the outbut buffer into the portaudio buffer
if (result && (out_ptr > out_end2))
@@ -622,6 +626,7 @@ int WavegenInitSound()

void WavegenInit(int rate, int wavemult_fact)
{
int ix;
double x;

if (wavemult_fact == 0)
@@ -639,7 +644,7 @@ void WavegenInit(int rate, int wavemult_fact)
wdata.amplitude = 32;
wdata.amplitude_fmt = 100;

for (int ix = 0; ix < N_EMBEDDED_VALUES; ix++)
for (ix = 0; ix < N_EMBEDDED_VALUES; ix++)
embedded_value[ix] = embedded_default[ix];

// set up window to generate a spread of harmonics from a
@@ -652,7 +657,7 @@ void WavegenInit(int rate, int wavemult_fact)
if (samplerate != 22050) {
// wavemult table has preset values for 22050 Hz, we only need to
// recalculate them if we have a different sample rate
for (int ix = 0; ix < wavemult_max; ix++) {
for (ix = 0; ix < wavemult_max; ix++) {
x = 127*(1.0 - cos(PI2*ix/wavemult_max));
wavemult[ix] = (int)x;
}
@@ -667,19 +672,24 @@ void WavegenInit(int rate, int wavemult_fact)

int GetAmplitude(void)
{
int amp;

// normal, none, reduced, moderate, strong
static const unsigned char amp_emphasis[5] = { 16, 16, 10, 16, 22 };

int amp = (embedded_value[EMBED_A])*55/100;
amp = (embedded_value[EMBED_A])*55/100;
general_amplitude = amp * amp_emphasis[embedded_value[EMBED_F]] / 16;
return general_amplitude;
}

static void WavegenSetEcho(void)
{
int delay;
int amp;

voicing = wvoice->voicing;
int delay = wvoice->echo_delay;
int amp = wvoice->echo_amp;
delay = wvoice->echo_delay;
amp = wvoice->echo_amp;

if (delay >= N_ECHO_BUF)
delay = N_ECHO_BUF-1;
@@ -726,15 +736,17 @@ int PeaksToHarmspect(wavegen_peaks_t *peaks, int pitch, int *htab, int control)
int fp; // centre freq of peak
int fhi; // high freq of peak
int h; // harmonic number
int pk;
int pk;
int hmax;
int hmax_samplerate; // highest harmonic allowed for the samplerate
int x;
int ix;
int ix;
int h1;

// initialise as much of *out as we will need
if (wvoice == NULL)
return 1;
int hmax = (peaks[wvoice->n_harmonic_peaks].freq + peaks[wvoice->n_harmonic_peaks].right)/pitch;
hmax = (peaks[wvoice->n_harmonic_peaks].freq + peaks[wvoice->n_harmonic_peaks].right)/pitch;
if (hmax >= MAX_HARMONIC)
hmax = MAX_HARMONIC-1;

@@ -763,9 +775,11 @@ int PeaksToHarmspect(wavegen_peaks_t *peaks, int pitch, int *htab, int control)
htab[h++] += pk_shape[(f-fp)/(p->right>>8)] * p->height;
}

int y;
int h2;
// increase bass
int y = peaks[1].height * 10; // addition as a multiple of 1/256s
int h2 = (1000<<16)/pitch; // decrease until 1000Hz
y = peaks[1].height * 10; // addition as a multiple of 1/256s
h2 = (1000<<16)/pitch; // decrease until 1000Hz
if (h2 > 0) {
x = y/h2;
h = 1;
@@ -801,7 +815,7 @@ int PeaksToHarmspect(wavegen_peaks_t *peaks, int pitch, int *htab, int control)
}

// adjust the amplitude of the first harmonic, affects tonal quality
int h1 = htab[1] * option_harmonic1;
h1 = htab[1] * option_harmonic1;
htab[1] = h1/8;

// calc intermediate increments of LF harmonics
@@ -817,13 +831,14 @@ static void AdvanceParameters()
{
// Called every 64 samples to increment the formant freq, height, and widths

int x;
int ix;
static int Flutter_ix = 0;

// advance the pitch
wdata.pitch_ix += wdata.pitch_inc;
if ((ix = wdata.pitch_ix>>8) > 127) ix = 127;
int x = wdata.pitch_env[ix] * wdata.pitch_range;
x = wdata.pitch_env[ix] * wdata.pitch_range;
wdata.pitch = (x>>8) + wdata.pitch_base;

amp_ix += amp_inc;
@@ -868,7 +883,9 @@ static void AdvanceParameters()

static double resonator(RESONATOR *r, double input)
{
double x = r->a * input + r->b * r->x1 + r->c * r->x2;
double x;

x = r->a * input + r->b * r->x1 + r->c * r->x2;
r->x2 = r->x1;
r->x1 = x;

@@ -881,13 +898,16 @@ static void setresonator(RESONATOR *rp, int freq, int bwidth, int init)
// bwidth Bandwidth of resonator in Hz
// init Initialize internal data

double x;
double arg;

if (init) {
rp->x1 = 0;
rp->x2 = 0;
}

double arg = minus_pi_t * bwidth;
double x = exp(arg);
arg = minus_pi_t * bwidth;
x = exp(arg);

rp->c = -(x * x);

@@ -899,19 +919,23 @@ static void setresonator(RESONATOR *rp, int freq, int bwidth, int init)

void InitBreath(void)
{
int ix;

minus_pi_t = -PI / samplerate;
two_pi_t = -2.0 * minus_pi_t;

for (int ix = 0; ix < N_PEAKS; ix++)
for (ix = 0; ix < N_PEAKS; ix++)
setresonator(&rbreath[ix], 2000, 200, 1);
}

static void SetBreath()
{
int pk;

if (wvoice->breath[0] == 0)
return;

for (int pk = 1; pk < N_PEAKS; pk++) {
for (pk = 1; pk < N_PEAKS; pk++) {
if (wvoice->breath[pk] != 0) {
// breath[0] indicates that some breath formants are needed
// set the freq from the current ynthesis formant and the width from the voice data
@@ -923,12 +947,14 @@ static void SetBreath()
static int ApplyBreath(void)
{
int value = 0;
int noise;
int ix;
int amp;

// use two random numbers, for alternate formants
int noise = (rand() & 0x3fff) - 0x2000;
noise = (rand() & 0x3fff) - 0x2000;

for (int ix = 1; ix < N_PEAKS; ix++) {
for (ix = 1; ix < N_PEAKS; ix++) {
if ((amp = wvoice->breath[ix]) != 0) {
amp *= (peaks[ix].height >> 14);
value += (int)resonator(&rbreath[ix], noise) * amp;
@@ -1256,6 +1282,7 @@ static int SetWithRange0(int value, int max)

static void SetPitchFormants()
{
int ix;
int factor = 256;
int pitch_value;

@@ -1268,7 +1295,7 @@ static void SetPitchFormants()
factor = 256 + (25 * (pitch_value - 50))/50;
}

for (int ix = 0; ix <= 5; ix++)
for (ix = 0; ix <= 5; ix++)
wvoice->freq[ix] = (wvoice->freq2[ix] * factor)/256;

factor = embedded_value[EMBED_T]*3;
@@ -1280,8 +1307,9 @@ void SetEmbedded(int control, int value)
{
// there was an embedded command in the text at this point
int sign = 0;
int command;

int command = control & 0x1f;
command = control & 0x1f;
if ((control & 0x60) == 0x60)
sign = -1;
else if ((control & 0x60) == 0x40)
@@ -1352,10 +1380,13 @@ static void SetAmplitude(int length, unsigned char *amp_env, int value)

void SetPitch2(voice_t *voice, int pitch1, int pitch2, int *pitch_base, int *pitch_range)
{
int x;
int base;
int range;
int pitch_value;

if (pitch1 > pitch2) {
int x = pitch1; // swap values
x = pitch1; // swap values
pitch1 = pitch2;
pitch2 = x;
}
@@ -1366,8 +1397,8 @@ void SetPitch2(voice_t *voice, int pitch1, int pitch2, int *pitch_base, int *pit
if (pitch_value < 0)
pitch_value = 0;

int base = (voice->pitch_base * pitch_adjust_tab[pitch_value])/128;
int range = (voice->pitch_range * embedded_value[EMBED_R])/50;
base = (voice->pitch_base * pitch_adjust_tab[pitch_value])/128;
range = (voice->pitch_range * embedded_value[EMBED_R])/50;

// compensate for change in pitch when the range is narrowed or widened
base -= (range - voice->pitch_range)*18;
@@ -1400,6 +1431,10 @@ void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v)
{
int ix;
DOUBLEX next;
int length2;
int length4;
int qix;
int cmd;
static int glottal_reduce_tab1[4] = { 0x30, 0x30, 0x40, 0x50 }; // vowel before [?], amp * 1/256
static int glottal_reduce_tab2[4] = { 0x90, 0xa0, 0xb0, 0xc0 }; // vowel after [?], amp * 1/256

@@ -1419,11 +1454,11 @@ void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v)
glottal_reduce = glottal_reduce_tab2[(modn >> 8) & 3];
}

for (int qix = wcmdq_head+1;; qix++) {
for (qix = wcmdq_head+1;; qix++) {
if (qix >= N_WCMDQ) qix = 0;
if (qix == wcmdq_tail) break;

int cmd = wcmdq[qix][0];
cmd = wcmdq[qix][0];
if (cmd == WCMD_SPECT) {
end_wave = 0; // next wave generation is from another spectrum
break;
@@ -1433,7 +1468,7 @@ void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v)
}

// round the length to a multiple of the stepsize
int length2 = (length + STEPSIZE/2) & ~0x3f;
length2 = (length + STEPSIZE/2) & ~0x3f;
if (length2 == 0)
length2 = STEPSIZE;

@@ -1441,7 +1476,7 @@ void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v)
samplecount_start = samplecount;
nsamples += length2;

int length4 = length2/4;
length4 = length2/4;

peaks[7].freq = (7800 * v->freq[7] + v->freqadd[7]*256) << 8;
peaks[8].freq = (9000 * v->freq[8] + v->freqadd[8]*256) << 8;
@@ -1487,7 +1522,9 @@ static int Wavegen2(int length, int modulation, int resume, frame_t *fr1, frame_
void Write4Bytes(FILE *f, int value)
{
// Write 4 bytes to a file, least significant first
for (int ix = 0; ix < 4; ix++) {
int ix;

for (ix = 0; ix < 4; ix++) {
fputc(value & 0xff, f);
value = value >> 8;
}
@@ -1642,12 +1679,13 @@ static int SpeedUp(short *outbuf, int length_in, int length_out, int end_of_text
// Call WavegenFill2, and then speed up the output samples.
int WavegenFill(int fill_zeros)
{
int finished;
unsigned char *p_start;

p_start = out_ptr;

// fill_zeros is ignored. It is now done in the portaudio callback
int finished = WavegenFill2(0);
finished = WavegenFill2(0);

#if HAVE_SONIC_H
if (sonicSpeed > 1.0) {

+ 17
- 9
src/speak-ng.c View File

@@ -127,6 +127,7 @@ USHORT voice_pcnt[N_PEAKS+1][3];

void DisplayVoices(FILE *f_out, char *language)
{
int ix;
const char *p;
int len;
int count;
@@ -153,7 +154,7 @@ void DisplayVoices(FILE *f_out, char *language)

fprintf(f_out, "Pty Language Age/Gender VoiceName File Other Languages\n");

for (int ix = 0; (v = voices[ix]) != NULL; ix++) {
for (ix = 0; (v = voices[ix]) != NULL; ix++) {
count = 0;
p = v->languages;
while (*p != 0) {
@@ -223,11 +224,13 @@ static int OpenWaveFile(const char *path, int rate)

static void CloseWaveFile()
{
unsigned int pos;

if ((f_wave == NULL) || (f_wave == stdout))
return;

fflush(f_wave);
unsigned int pos = ftell(f_wave);
pos = ftell(f_wave);

fseek(f_wave, 4, SEEK_SET);
Write4Bytes(f_wave, pos - 8);
@@ -235,19 +238,21 @@ static void CloseWaveFile()
fseek(f_wave, 40, SEEK_SET);
Write4Bytes(f_wave, pos - 44);


fclose(f_wave);
f_wave = NULL;
}

static int WavegenFile(void)
{
int finished;
unsigned char wav_outbuf[1024];
char fname[210];

out_ptr = out_start = wav_outbuf;
out_end = wav_outbuf + sizeof(wav_outbuf);

int finished = WavegenFill(0);
finished = WavegenFill(0);

if (quiet)
return finished;
@@ -329,6 +334,8 @@ static void init_path(char *argv0, char *path_specified)

static int initialise(void)
{
int param;
int result;
int srate = 22050; // default sample rate

// It seems that the wctype functions don't work until the locale has been set
@@ -339,7 +346,6 @@ static int initialise(void)
setlocale(LC_CTYPE, "");
}

int result;
if ((result = LoadPhData(&srate)) != 1) {
if (result == -1) {
fprintf(stderr, "Failed to load espeak-data\n");
@@ -352,7 +358,7 @@ static int initialise(void)
SetVoiceStack(NULL, "");
SynthesizeInit();

for (int param = 0; param < N_SPEECH_PARAM; param++)
for (param = 0; param < N_SPEECH_PARAM; param++)
param_stack[0].parameter[param] = param_defaults[param];

return 0;
@@ -405,6 +411,7 @@ int main(int argc, char **argv)
int c;
int value;
int speed = 175;
int ix;
char *optarg2;
int amp = 100; // default
int wordgap = 0;
@@ -435,6 +442,7 @@ int main(int argc, char **argv)
optind = 1;
opt_string = "";
while (optind < argc) {
int len;
char *p;

if ((c = *opt_string) == 0) {
@@ -454,10 +462,10 @@ int main(int argc, char **argv)
break; // -- means don't interpret further - as commands

opt_string = "";
for (int ix = 0;; ix++) {
for (ix = 0;; ix++) {
if (long_options[ix].name == 0)
break;
size_t len = strlen(long_options[ix].name);
len = strlen(long_options[ix].name);
if (memcmp(long_options[ix].name, p, len) == 0) {
c = long_options[ix].val;
optarg2 = NULL;
@@ -559,7 +567,7 @@ int main(int argc, char **argv)
case 0x103: // --punct
option_punctuation = 1;
if (optarg2 != NULL) {
int ix = 0;
ix = 0;
while ((ix < N_PUNCTLIST) && ((option_punctlist[ix] = optarg2[ix]) != 0)) ix++;
option_punctlist[N_PUNCTLIST-1] = 0;
option_punctuation = 2;
@@ -726,7 +734,7 @@ int main(int argc, char **argv)
InitText(0);
SpeakNextClause(f_text, p_text, 0);

int ix = 1;
ix = 1;
for (;;) {
if (WavegenFile() != 0) {
if (ix == 0)

Write Preview
Loading…
Cancel
Save