ref: e98804740dd0d36ee04d710dc0859d31aa0126a5
parent: b273a40041a0330bd1545316eed63959a6cc4bdd
author: Alan W Black <awb@cs.cmu.edu>
date: Mon Jan 21 04:41:21 EST 2019
post-wilderness fixes
--- /dev/null
+++ b/.time-stamp
@@ -1,0 +1,8 @@
+Festival Lite
+flite
+2.2
+Sep 2018
+current
+awb
+leith
+Tue Sep 11 08:19:34 EDT 2018
--- a/Makefile
+++ b/Makefile
@@ -37,8 +37,8 @@
## Authors: Alan W Black (awb@cs.cmu.edu) ##
## Kevin A. Lenzo (lenzo@cs.cmu.edu) ##
## and others see ACKNOWLEDGEMENTS ##
-## Date: Dec 2017 ##
-## Version: 2.1 release ##
+## Date: Sep 2018 ##
+## Version: 2.2 current ##
## ##
###########################################################################
TOP=.
--- a/README.md
+++ b/README.md
@@ -1,7 +1,7 @@
Flite: a small run-time speech synthesis engine
version 2.1-release
- Copyright Carnegie Mellon University 1999-2017
+ Copyright Carnegie Mellon University 1999-2018
All rights reserved
http://cmuflite.org
https://github.com/festvox/flite
@@ -137,6 +137,9 @@
News
----
+New in 2.2 (Oct 2018)
+ o Better qrapheme support (Wilderness Languages) hundreds of new languages
+
New in 2.1 (Oct 2017)
o Improved Indic front end support (thanks to Suresh Bazaj @ Hear2Read)
--- a/config/project.mak
+++ b/config/project.mak
@@ -36,8 +36,8 @@
###########################################################################
PROJECT_NAME = Festival Lite
PROJECT_PREFIX = flite
-PROJECT_VERSION = 2.1
-PROJECT_DATE = Dec 2017
-PROJECT_STATE = release
+PROJECT_VERSION = 2.2
+PROJECT_DATE = Sep 2018
+PROJECT_STATE = current
PROJECT_SHLIB_VERSION = 1
--- a/include/cst_cg.h
+++ b/include/cst_cg.h
@@ -51,9 +51,10 @@
#include "cst_wave.h"
#include "cst_audio.h"
#include "cst_synth.h" /* for dur_stat */
+#include "cst_phoneset.h"
/* The parameters in the spectral model(s) may be encoded in various ways */
-/* mostly to preserve space. The decoded can be dependent on the shape type */
+/* mostly to preserve space. The decoder can be dependent on the shape type */
/* This needs to be coordinated with the model creation in make_cg.scm */
/* This is currently very BIG_HAMMER-like and hard coded */
@@ -142,6 +143,11 @@
int spamf0;
float gain;
+
+ /* If its a "grapheme" voice there will be a phoneset and a char_map */
+ const cst_phoneset *phoneset;
+ const char * const *char_sym_map; /* Unicode char to symbol map */
+ const char * const *sym_phone_map; /* symbol to phone map */
int freeable; /* doesn't get dumped, but 1 when this a freeable struct */
--- a/include/flite_version.h
+++ b/include/flite_version.h
@@ -1,6 +1,6 @@
#define FLITE_PROJECT_NAME "Festival Lite"
#define FLITE_PROJECT_PREFIX "flite"
-#define FLITE_PROJECT_VERSION "2.0.7"
-#define FLITE_PROJECT_DATE "Jul 2017"
+#define FLITE_PROJECT_VERSION "2.2"
+#define FLITE_PROJECT_DATE "Sep 2018"
#define FLITE_PROJECT_STATE "current"
#define FLITE_PROJECT_SHLIB_VERSION "1"
--- a/lang/cmu_grapheme_lex/cmu_grapheme_lex.c
+++ b/lang/cmu_grapheme_lex/cmu_grapheme_lex.c
@@ -89,6 +89,7 @@
/* afterwards */
cst_utf8_as_hex(val_string(val_car(v)),ord);
phindex = cst_find_u2sampa(ord);
+ printf("awb_debug lookup sampa %s %s\n",val_string(val_car(v)),ord);#if 0
if (phindex < 0)
printf("awb_debug no sampa %s %s\n",val_string(val_car(v)),ord);--- a/lang/cmu_indic_lang/indic_eng_num_table.h
+++ b/lang/cmu_indic_lang/indic_eng_num_table.h
@@ -42,8 +42,133 @@
#include "indic_num_table.h"
+#if 0
static const char * const indic_eng_digit[11][2] =
{+ { "0", "शून्य" },+ { "1", "एक" },+ { "2", "दोन" },+ { "3", "तीन" },+ { "4", "चार" },+ { "5", "पाच" },+ { "6", "सहा" },+ { "7", "सात" },+ { "8", "आठ" },+ { "9", "नऊ" },+ { NULL, NULL },+};
+
+static const char * const indic_eng_two_digit[101][4] =
+{+ { "1", "0", "दहा", NULL },+ { "1", "1", "अकरा", NULL },+ { "1", "2", "बारा", NULL },+ { "1", "3", "तेरा", NULL },+ { "1", "4", "चौदा", NULL },+ { "1", "5", "पंधरा", NULL },+ { "1", "6", "सोळा", NULL },+ { "1", "7", "सतरा", NULL },+ { "1", "8", "अठरा", NULL },+ { "1", "9", "एकोणीस", NULL },+ { "2", "0", "वीस", NULL },+ { "2", "1", "एकवीस", NULL },+ { "2", "2", "बावीस", NULL },+ { "2", "3", "तेवीस", NULL },+ { "2", "4", "चोवीस", NULL },+ { "2", "5", "पंचवीस", NULL },+ { "2", "6", "सव्वीस", NULL },+ { "2", "7", "सत्तावीस", NULL },+ { "2", "8", "अठ्ठावीस", NULL },+ { "2", "9", "एकोणतीस", NULL },+ { "3", "0", "तीस", NULL },+ { "3", "1", "एकतीस", NULL },+ { "3", "2", "बत्तीस", NULL },+ { "3", "3", "तेहेतीस", NULL },+ { "3", "4", "चौतीस", NULL },+ { "3", "5", "पस्तीस", NULL },+ { "3", "6", "छत्तीस", NULL },+ { "3", "7", "सदतीस", NULL },+ { "3", "8", "अडतीस", NULL },+ { "3", "9", "एकोणचाळीस", NULL },+ { "4", "0", "चाळीस", NULL },+ { "4", "1", "एक्केचाळीस", NULL },+ { "4", "2", "बेचाळीस", NULL },+ { "4", "3", "त्रेचाळीस", NULL },+ { "4", "4", "चव्वेचाळीस", NULL },+ { "4", "5", "पंचेचाळीस", NULL },+ { "4", "6", "सेहेचाळीस", NULL },+ { "4", "7", "सत्तेचाळीस", NULL },+ { "4", "8", "अठ्ठेचाळीस", NULL },+ { "4", "9", "एकोणपन्नास", NULL },+ { "5", "0", "पन्नास", NULL },+ { "5", "1", "एक्कावन्न", NULL },+ { "5", "2", "बावन्न", NULL },+ { "5", "3", "त्रेपन्न", NULL },+ { "5", "4", "चोपन्न", NULL },+ { "5", "5", "पंचावन्न", NULL },+ { "5", "6", "छप्पन्न", NULL },+ { "5", "7", "सत्तावन्न", NULL },+ { "5", "8", "अठ्ठावन्न", NULL },+ { "5", "9", "एकोणसाठ", NULL },+ { "6", "0", "साठ", NULL },+ { "6", "1", "एकसष्ठ", NULL },+ { "6", "2", "बासष्ठ", NULL },+ { "6", "3", "त्रेसष्ठ", NULL },+ { "6", "4", "चौसष्ठ", NULL },+ { "6", "5", "पासष्ठ", NULL },+ { "6", "6", "सहासष्ठ", NULL },+ { "6", "7", "सदुसष्ठ", NULL },+ { "6", "8", "अडुसष्ठ", NULL },+ { "6", "9", "एकोणसत्तर", NULL },+ { "7", "0", "सत्तर", NULL },+ { "7", "1", "एक्काहत्तर", NULL },+ { "7", "2", "बाहत्तर", NULL },+ { "7", "3", "त्र्याहत्तर", NULL },+ { "7", "4", "चौर्याहत्तर", NULL },+ { "7", "5", "पंच्याहत्तर", NULL },+ { "7", "6", "शहात्तर", NULL },+ { "7", "7", "सत्याहत्तर", NULL },+ { "7", "8", "अठ्ठ्याहत्तर", NULL },+ { "7", "9", "एकोणऐंशी", NULL },+ { "8", "0", "ऐंशी", NULL },+ { "8", "1", "एक्क्याऐंशी", NULL },+ { "8", "2", "ब्याऐंशी", NULL },+ { "8", "3", "त्र्याऐंशी", NULL },+ { "8", "4", "चौऱ्याऐंशी", NULL },+ { "8", "5", "पंच्याऐंशी", NULL },+ { "8", "6", "शहाऐंशी", NULL },+ { "8", "7", "सत्त्याऐंशी", NULL },+ { "8", "8", "अठ्ठ्याऐंशी", NULL },+ { "8", "9", "एकोणनव्वद", NULL },+ { "9", "0", "नव्वद", NULL },+ { "9", "1", "एक्क्याण्णव", NULL },+ { "9", "2", "ब्याण्णव", NULL },+ { "9", "3", "त्र्याण्णव", NULL },+ { "9", "4", "चौऱ्याण्णव", NULL },+ { "9", "5", "पंच्याण्णव", NULL },+ { "9", "6", "शहाण्णव", NULL },+ { "9", "7", "सत्त्याण्णव", NULL },+ { "9", "8", "अठ्ठ्याण्णव", NULL },+ { "9", "9", "नव्व्याण्णव", NULL },+ { NULL, NULL },+};
+
+const static indic_num_table mar_num_table = {+ "mar",
+ &indic_eng_digit,
+ &indic_eng_two_digit,
+ "शंभर", /* hundred */
+ "हजार", /* thousand */
+ "लाख", /* lakh */
+ "कोटी", /* crore */
+};
+#endif
+
+
+#if 1
+
+static const char * const indic_eng_digit[11][2] =
+{ { "0", "zero" }, { "1", "one" }, { "2", "two" },@@ -171,6 +296,8 @@
"lakh",
"crore"
};
+
+#endif
#ifdef __cplusplus
--- a/src/cg/cst_cg.c
+++ b/src/cg/cst_cg.c
@@ -367,6 +367,8 @@
start_index = ffeature_int(syl,"R:SylStructure.daughter1.R:segstate.daughter1.R:mcep_link.daughter1.frame_number");
end_index = ffeature_int(syl,"R:SylStructure.daughtern.R:segstate.daughtern.R:mcep_link.daughtern.frame_number");
mid_index = (int)((start_index + end_index)/2.0);
+ if (end_index <= start_index)
+ continue;
start_f0 = param_track->frames[start_index][0];
if (end_f0 > 0.0)
--- a/src/cg/cst_cg_map.c
+++ b/src/cg/cst_cg_map.c
@@ -176,6 +176,8 @@
db->spamf0 = cst_read_int(fd,bs); /* yes, twice, its above too */
db->gain = cst_read_float(fd,bs);
+ /* If this is "grapheme" voice, we will have phoneset and char_map */
+
return db;
}
--- a/testsuite/lex_lookup_main.c
+++ b/testsuite/lex_lookup_main.c
@@ -50,6 +50,12 @@
p = lex_lookup(l,word,pos,NULL);
val_print(stdout,p);
printf("\n");+ if (l->lts_function)
+ p = l->lts_function(l,word,pos,NULL);
+ else if (l->lts_rule_set)
+ p = lts_apply(word,"",l->lts_rule_set);
+ val_print(stdout,p);
+ printf("\n");delete_val(p);
}
--- a/tools/Makefile.flite
+++ b/tools/Makefile.flite
@@ -67,6 +67,9 @@
ifeq ($(SPAMF0),true)
CSRCS += $(VOICENAME)_spamf0_phrase.c $(VOICENAME)_spamf0_accent.c $(VOICENAME)_spamf0_accent_params.c
endif
+ifeq ($(CG_GRAPHEME),true)
+CSRCS += $(VOICENAME)_phoneset.c
+endif
MCEPOBJS =
LPCOBJS =
else
--- a/tools/VOICE_cg.c
+++ b/tools/VOICE_cg.c
@@ -48,6 +48,8 @@
extern cst_cg_db __VOICENAME___cg_db;
cst_voice *__VOICENAME___cg = NULL;
+extern const cst_phoneset __VOICENAME___phoneset;
+
cst_voice *register___VOICENAME__(const char *voxdir)
{cst_voice *vox;
@@ -72,6 +74,12 @@
lex = __FLITELEX___init();
flite_feat_set(vox->features,"lexicon",lexicon_val(lex));
flite_feat_set(vox->features,"postlex_func",uttfunc_val(lex->postlex));
+
+ /* If grapheme link in grapheme specific info */
+ if (feat_present(vox->features,"grapheme"))
+ {+ flite_feat_set(vox->features,"phoneset",phoneset_val(&__VOICENAME___phoneset));
+ }
/* No standard segment durations are needed as its done at the */
/* HMM state level */
--- a/tools/build_flite
+++ b/tools/build_flite
@@ -387,12 +387,15 @@
then
${FLITEDIR}/tools/make_flite_feats etc/voice.feats >flite/${FV_VOICENAME}_voice_feats.cfi
-
- if [ "$FV_LANG_TYPE" = "grapheme" ]
+
+ # If Grapheme-based it has its own phoneset
+ if [ -f festvox/${FV_VOICENAME}_char_phone_map.scm ]then
- echo cg_convert: converting phoneset table
+ echo "flite_feat_set_int(vox->features,\"grapheme\",1);" >>flite/${FV_VOICENAME}_voice_feats.c+ echo cg_convert: converting phoneset table
+ echo "CG_GRAPHEME=true" >>flite/paramfiles.mak
$FESTIVAL -b $FLITEDIR/tools/make_phoneset.scm \
- '(phonesettoC "'${FV_INST}_${FV_LANG}'" (car (load "festvox/'${FV_VOICENAME}'_phoneset.scm" t)) "pau" "flite")'+ '(phonesettoC "'${FV_VOICENAME}'" (car (load "festvox/'${FV_VOICENAME}'_phoneset.scm" t)) "pau" "flite")'fi
echo "flite_build cg complete. You can compile the generated voice by"
--- a/tools/make_cg.scm
+++ b/tools/make_cg.scm
@@ -75,6 +75,8 @@
(format ofd "#include \"cst_cg.h\"\n")
(format ofd "#include \"cst_cart.h\"\n")
+ (format ofd "extern const cst_phoneset %s_phoneset;\n\n" name)
+
(format t "cg_convert: converting F0 trees\n")
;; F0 trees
(if (and cg:rfs_models (probe_file "rf_models/mlistf0"))
@@ -171,6 +173,20 @@
(format ofd "extern const unsigned short * const %s_01_model_vectors[];\n" name )
))
+ (if (probe_file (format nil "festvox/%s_char_phone_map.scm" name))
+ (begin
+ (setq cpm (load (format nil "festvox/%s_char_phone_map.scm" name) t))
+ (format ofd "\nstatic char * const %s_char_phone_map[%s][2] =\n"
+ name (+ 1 (length cpm)))
+ (format ofd "{\n")+ (mapcar
+ (lambda (x)
+ (format ofd " { %l, \"%s\" },\n" (car x) (cadr x)))+ cpm)
+ (format ofd " { NULL, NULL }\n")+ (format ofd "};\n\n")
+ ))
+
(format ofd "#define %s_num_f0_models %d\n" name (length f0ms))
(format ofd "const cst_cart **%s_f0_carts_table[] = {\n" name)(mapcar
@@ -399,9 +415,18 @@
(format ofd " 0,0, /* cg:mixed_excitation */\n")
(format ofd " NULL, \n")))
(if cg:spamf0
- (format ofd " 1, // cg:spamf0\n")
- (format ofd " 0, // cg:spamf0\n"))
+ (format ofd " 1, /* cg:spamf0 */\n")
+ (format ofd " 0, /* cg:spamf0 */\n"))
(format ofd " 1.5 /* gain */\n")
+
+ ;; If a grapheme language, add phoneset and char_phone_map
+ (if (probe_file (format nil "festvox/%s_char_phone_map.scm" name))
+ (begin
+ (format ofd " ,\n")
+ (format ofd " &%s_phoneset,\n" name)
+ (format ofd " &%s_char_phone_map\n" name)
+ ))
+
(format ofd "};\n")
(fclose ofd)
@@ -802,6 +827,30 @@
(string-before x "}")
"rb"
(string-after x "}"))))
+ ((string-matches x ".*].*")
+ (intern
+ (string-append
+ (string-before x "]")
+ "rbk"
+ (string-after x "]"))))
+ ((string-matches x ".*-.*")
+ (intern
+ (string-append
+ (string-before x "-")
+ "hyp"
+ (string-after x "-"))))
+ ((string-matches x ".*\\*.*")
+ (intern
+ (string-append
+ (string-before x "*")
+ "star"
+ (string-after x "*"))))
+ ((string-matches x ".*\\^.*")
+ (intern
+ (string-append
+ (string-before x "^")
+ "caret"
+ (string-after x "^"))))
((string-matches x ".*~.*")
(intern
(string-append