mjohnson/espeak-ng
1
0
Fork 0
mirror of https://github.com/rhasspy/espeak-ng.git synced 2026-10-31 05:35:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #795

Browse Source
This commit is contained in:
Valdis Vitolins
2020-07-22 16:27:09 +03:00
parent 02174e2f76 5d03047c19
commit bddd8a8bde
9 changed files with 43 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/libespeak-ng/klatt.c
+25 -32
View File
@@ -40,9 +40,7 @@
#include "klatt.h"
extern unsigned char *out_ptr;
extern unsigned char *out_start;
extern unsigned char *out_end;
extern WGEN_DATA wdata;
static int nsamples;
static int sample_count;
@@ -63,6 +61,7 @@ static double gen_noise(double);
static double DBtoLIN(long);
static void frame_init(klatt_frame_ptr);
static void setabc(long, long, resonator_ptr);
static void SetSynth_Klatt(int length, frame_t *fr1, frame_t *fr2, voice_t *v, int control);
static void setzeroabc(long, long, resonator_ptr);
static klatt_frame_t kt_frame;
@@ -242,7 +241,7 @@ static double sampled_source(int source_num)
Converts synthesis parameters to a waveform.
*/
static int parwave(klatt_frame_ptr frame)
static int parwave(klatt_frame_ptr frame, WGEN_DATA *wdata)
{
double temp;
int value;
@@ -376,24 +375,24 @@ static int parwave(klatt_frame_ptr frame)
out = outbypas - out;
out = resonator(&(kt_globals.rsn[Rout]), out);
temp = (int)(out * wdata.amplitude * kt_globals.amp_gain0); // Convert back to integer
temp = (int)(out * wdata->amplitude * kt_globals.amp_gain0); // Convert back to integer
// mix with a recorded WAV if required for this phoneme
signed char c;
int sample;
if (wdata.mix_wavefile_ix < wdata.n_mix_wavefile) {
if (wdata.mix_wave_scale == 0) {
if (wdata->mix_wavefile_ix < wdata->n_mix_wavefile) {
if (wdata->mix_wave_scale == 0) {
// a 16 bit sample
c = wdata.mix_wavefile[wdata.mix_wavefile_ix+1];
sample = wdata.mix_wavefile[wdata.mix_wavefile_ix] + (c * 256);
wdata.mix_wavefile_ix += 2;
c = wdata->mix_wavefile[wdata->mix_wavefile_ix+1];
sample = wdata->mix_wavefile[wdata->mix_wavefile_ix] + (c * 256);
wdata->mix_wavefile_ix += 2;
} else {
// a 8 bit sample, scaled
sample = (signed char)wdata.mix_wavefile[wdata.mix_wavefile_ix++] * wdata.mix_wave_scale;
sample = (signed char)wdata->mix_wavefile[wdata->mix_wavefile_ix++] * wdata->mix_wave_scale;
}
int z2 = sample * wdata.amplitude_v / 1024;
z2 = (z2 * wdata.mix_wave_amp)/40;
int z2 = sample * wdata->amplitude_v / 1024;
z2 = (z2 * wdata->mix_wave_amp)/40;
temp += z2;
}
@@ -838,15 +837,17 @@ static double DBtoLIN(long dB)
return (double)(amptable[dB]) * 0.001;
}
extern voice_t *wvoice;
static klatt_peaks_t peaks[N_PEAKS];
static int end_wave;
static int klattp[N_KLATTP];
static double klattp1[N_KLATTP];
static double klattp_inc[N_KLATTP];
static int Wavegen_Klatt(int resume)
int Wavegen_Klatt(int length, int resume, frame_t *fr1, frame_t *fr2, WGEN_DATA *wdata, voice_t *wvoice)
{
if (resume == 0)
SetSynth_Klatt(length, fr1, fr2, wvoice, 1);
int pk;
int x;
int ix;
@@ -856,7 +857,7 @@ static int Wavegen_Klatt(int resume)
sample_count = 0;
while (sample_count < nsamples) {
kt_frame.F0hz10 = (wdata.pitch * 10) / 4096;
kt_frame.F0hz10 = (wdata->pitch * 10) / 4096;
// formants F6,F7,F8 are fixed values for cascade resonators, set in KlattInit()
// but F6 is used for parallel resonator
@@ -904,10 +905,10 @@ static int Wavegen_Klatt(int resume)
}
// advance the pitch
wdata.pitch_ix += wdata.pitch_inc;
if ((ix = wdata.pitch_ix>>8) > 127) ix = 127;
x = wdata.pitch_env[ix] * wdata.pitch_range;
wdata.pitch = (x>>8) + wdata.pitch_base;
wdata->pitch_ix += wdata->pitch_inc;
if ((ix = wdata->pitch_ix>>8) > 127) ix = 127;
x = wdata->pitch_env[ix] * wdata->pitch_range;
wdata->pitch = (x>>8) + wdata->pitch_base;
kt_globals.nspfr = (nsamples - sample_count);
if (kt_globals.nspfr > STEPSIZE)
@@ -915,7 +916,7 @@ static int Wavegen_Klatt(int resume)
frame_init(&kt_frame); // get parameters for next frame of speech
if (parwave(&kt_frame) == 1)
if (parwave(&kt_frame, wdata) == 1)
return 1; // output buffer is full
}
@@ -927,14 +928,14 @@ static int Wavegen_Klatt(int resume)
end_wave = 0;
sample_count -= fade;
kt_globals.nspfr = fade;
if (parwave(&kt_frame) == 1)
if (parwave(&kt_frame, wdata) == 1)
return 1; // output buffer is full
}
return 0;
}
static void SetSynth_Klatt(int length, frame_t *fr1, frame_t *fr2, voice_t *v, int control)
static void SetSynth_Klatt(int length, frame_t *fr1, frame_t *fr2, voice_t *wvoice, int control)
{
int ix;
double next;
@@ -1004,9 +1005,9 @@ static void SetSynth_Klatt(int length, frame_t *fr1, frame_t *fr2, voice_t *v, i
nsamples = length;
for (ix = 1; ix < 6; ix++) {
peaks[ix].freq1 = (fr1->ffreq[ix] * v->freq[ix] / 256.0) + v->freqadd[ix];
peaks[ix].freq1 = (fr1->ffreq[ix] * wvoice->freq[ix] / 256.0) + wvoice->freqadd[ix];
peaks[ix].freq = (int)peaks[ix].freq1;
next = (fr2->ffreq[ix] * v->freq[ix] / 256.0) + v->freqadd[ix];
next = (fr2->ffreq[ix] * wvoice->freq[ix] / 256.0) + wvoice->freqadd[ix];
peaks[ix].freq_inc = ((next - peaks[ix].freq1) * STEPSIZE) / length;
if (ix < 4) {
@@ -1050,14 +1051,6 @@ static void SetSynth_Klatt(int length, frame_t *fr1, frame_t *fr2, voice_t *v, i
}
}
int Wavegen_Klatt2(int length, int resume, frame_t *fr1, frame_t *fr2)
{
if (resume == 0)
SetSynth_Klatt(length, fr1, fr2, wvoice, 1);
return Wavegen_Klatt(resume);
}
void KlattInit()
{
src/libespeak-ng/klatt.h
+1 -2
View File
@@ -179,8 +179,7 @@ typedef struct {
void KlattInit(void);
void KlattReset(int control);
int Wavegen_Klatt2(int length, int resume, frame_t *fr1, frame_t *fr2);
int Wavegen_Klatt(int length, int resume, frame_t *fr1, frame_t *fr2, WGEN_DATA *wdata, voice_t *wvoice);
#ifdef __cplusplus
}
src/libespeak-ng/phoneme.h
-2
View File
@@ -249,9 +249,7 @@ phoneme_add_feature(PHONEME_TAB *phoneme,
// Several phoneme tables may be loaded into memory. phoneme_tab points to
// one for the current voice
extern int n_phoneme_tab;
extern int current_phoneme_table;
extern PHONEME_TAB *phoneme_tab[N_PHONEME_TAB];
extern unsigned char phoneme_tab_flags[N_PHONEME_TAB]; // bit 0: not inherited
typedef struct {
char name[N_PHONEME_TAB_NAME];
src/libespeak-ng/speech.c
+1
View File
@@ -65,6 +65,7 @@
unsigned char *outbuf = NULL;
int outbuf_size = 0;
unsigned char *out_start;
espeak_EVENT *event_list = NULL;
int event_list_ix = 0;
src/libespeak-ng/synthdata.c
+5 -14
View File
@@ -45,9 +45,7 @@ const int version_phdata = 0x014801;
// copy the current phoneme table into here
int n_phoneme_tab;
int current_phoneme_table;
PHONEME_TAB *phoneme_tab[N_PHONEME_TAB];
unsigned char phoneme_tab_flags[N_PHONEME_TAB]; // bit 0: not inherited
unsigned short *phoneme_index = NULL;
char *phondata_ptr = NULL;
@@ -62,7 +60,6 @@ int wavefile_ix; // a wavefile to play along with the synthesis
int wavefile_amp;
int seq_len_adjust;
int vowel_transition[4];
static espeak_ng_STATUS ReadPhFile(void **ptr, const char *fname, int *size, espeak_ng_ERROR_CONTEXT *context)
{
@@ -331,19 +328,16 @@ unsigned char *GetEnvelope(int index)
return (unsigned char *)&phondata_ptr[index];
}
static void SetUpPhonemeTable(int number, bool recursing)
static void SetUpPhonemeTable(int number)
{
int ix;
int includes;
int ph_code;
PHONEME_TAB *phtab;
if (recursing == false)
memset(phoneme_tab_flags, 0, sizeof(phoneme_tab_flags));
if ((includes = phoneme_tab_list[number].includes) > 0) {
// recursively include base phoneme tables
SetUpPhonemeTable(includes-1, true);
SetUpPhonemeTable(includes - 1);
}
// now add the phonemes from this table
@@ -353,18 +347,15 @@ static void SetUpPhonemeTable(int number, bool recursing)
phoneme_tab[ph_code] = &phtab[ix];
if (ph_code > n_phoneme_tab)
n_phoneme_tab = ph_code;
if (recursing == 0)
phoneme_tab_flags[ph_code] |= 1; // not inherited
}
}
void SelectPhonemeTable(int number)
int SelectPhonemeTable(int number)
{
n_phoneme_tab = 0;
SetUpPhonemeTable(number, false); // recursively for included phoneme tables
SetUpPhonemeTable(number); // recursively for included phoneme tables
n_phoneme_tab++;
current_phoneme_table = number;
return number;
}
int LookupPhonemeTable(const char *name)
src/libespeak-ng/synthdata.h
+1 -1
View File
@@ -52,7 +52,7 @@ frameref_t *LookupSpect(PHONEME_TAB *this_ph,
int NumInstnWords(unsigned short *prog);
int PhonemeCode(unsigned int mnem);
void SelectPhonemeTable(int number);
int SelectPhonemeTable(int number);
int SelectPhonemeTableName(const char *name);
#ifdef __cplusplus
src/libespeak-ng/synthesize.c
+2 -4
View File
@@ -44,7 +44,6 @@
#include "synthesize.h"
#include "translate.h"
extern FILE *f_log;
static void SmoothSpect(void);
// list of phonemes in a clause
@@ -127,7 +126,6 @@ static void EndPitch(int voice_break)
syllable_end = wcmdq_tail;
SmoothSpect();
syllable_centre = -1;
memset(vowel_transition, 0, sizeof(vowel_transition));
}
}
@@ -1190,7 +1188,6 @@ int Generate(PHONEME_LIST *phoneme_list, int *n_ph, bool resume)
syllable_end = wcmdq_tail;
syllable_centre = -1;
last_pitch_cmd = -1;
memset(vowel_transition, 0, sizeof(vowel_transition));
memset(&worddata, 0, sizeof(worddata));
DoPause(0, 0); // isolate from the previous clause
}
@@ -1549,6 +1546,7 @@ int Generate(PHONEME_LIST *phoneme_list, int *n_ph, bool resume)
return 0; // finished the phoneme list
}
static int current_phoneme_table;
int SpeakNextClause(int control)
{
// Speak text from memory (text_in)
@@ -1577,7 +1575,7 @@ int SpeakNextClause(int control)
}
if (current_phoneme_table != voice->phoneme_tab_ix)
SelectPhonemeTable(voice->phoneme_tab_ix);
current_phoneme_table = SelectPhonemeTable(voice->phoneme_tab_ix);
// read the next clause from the input text file, translate it, and generate
// entries in the wavegen command queue
src/libespeak-ng/synthesize.h
-2
View File
@@ -450,7 +450,6 @@ extern int samplerate_native;
extern int wavefile_ix;
extern int wavefile_amp;
extern int vowel_transition[4];
#define N_ECHO_BUF 5500 // max of 250mS at 22050 Hz
extern int echo_head;
@@ -482,7 +481,6 @@ extern int formant_rate[]; // max rate of change of each formant
extern SPEED_FACTORS speed;
extern unsigned char *out_ptr;
extern unsigned char *out_start;
extern unsigned char *out_end;
extern espeak_EVENT *event_list;
extern t_espeak_callback *synth_callback;
src/libespeak-ng/wavegen.c
+8 -13
View File
@@ -51,9 +51,10 @@
#define N_WAV_BUF 10
static void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *v);
voice_t *wvoice = NULL;
FILE *f_log = NULL;
static int option_harmonic1 = 10;
static int flutter_amp = 64;
@@ -113,7 +114,6 @@ static double minus_pi_t;
static double two_pi_t;
unsigned char *out_ptr;
unsigned char *out_start;
unsigned char *out_end;
espeak_ng_OUTPUT_HOOKS* output_hooks = NULL;
@@ -680,8 +680,11 @@ static int ApplyBreath(void)
return value;
}
static int Wavegen()
static int Wavegen(int length, int modulation, bool resume, frame_t *fr1, frame_t *fr2, voice_t *wvoice)
{
if (resume == false)
SetSynth(length, modulation, fr1, fr2, wvoice);
if (wvoice == NULL)
return 0;
@@ -1245,14 +1248,6 @@ static void SetSynth(int length, int modn, frame_t *fr1, frame_t *fr2, voice_t *
}
}
static int Wavegen2(int length, int modulation, bool resume, frame_t *fr1, frame_t *fr2)
{
if (resume == false)
SetSynth(length, modulation, fr1, fr2, wvoice);
return Wavegen();
}
void Write4Bytes(FILE *f, int value)
{
// Write 4 bytes to a file, least significant first
@@ -1339,14 +1334,14 @@ static int WavegenFill2()
wdata.n_mix_wavefile = 0; // ... and drop through to WCMD_SPECT case
case WCMD_SPECT:
echo_complete = echo_length;
result = Wavegen2(length & 0xffff, q[1] >> 16, resume, (frame_t *)q[2], (frame_t *)q[3]);
result = Wavegen(length & 0xffff, q[1] >> 16, resume, (frame_t *)q[2], (frame_t *)q[3], wvoice);
break;
#ifdef INCLUDE_KLATT
case WCMD_KLATT2: // as WCMD_SPECT but stop any concurrent wave file
wdata.n_mix_wavefile = 0; // ... and drop through to WCMD_SPECT case
case WCMD_KLATT:
echo_complete = echo_length;
result = Wavegen_Klatt2(length & 0xffff, resume, (frame_t *)q[2], (frame_t *)q[3]);
result = Wavegen_Klatt(length & 0xffff, resume, (frame_t *)q[2], (frame_t *)q[3], &wdata, wvoice);
break;
#endif
case WCMD_MARKER: