File: NEWS

package info (click to toggle)
festival 1:2.5.0-3
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 7,508 kB
  • sloc: cpp: 27,915; lisp: 15,725; ansic: 6,022; sh: 5,683; java: 1,536; makefile: 753; xml: 291; perl: 87
file content (219 lines) | stat: -rw-r--r-- 8,538 bytes parent folder | download | duplicates (10)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219

Note that not all features discussed in this file are included in
the standard distribution.

HISTORY

June 21st 2001 1.4.2 Release
   Various new gcc's support
   Visual C++ 6.0 support
   uses configure (though could do so even more)
   substantial updates to the clunits unit selection module
   lots of wee bugs fixed
   a few very hard bugs fixed 
     (client/server race condition)
     (dropped bytes in reading files when machine overloaded)
     (FreeBSD memory/gc problem)
   default waverform type is now RIFF.

Nov 21st 1999 1.4.1 Release
   SSFF (for emulabel) track support
   AIX support
   Java fixes
   various minor bug fixes
   WFST with proper quoting
   Wagon sample counts
   gcc-2.95.1 support

June 20th 1999 1.4.0 Release 
   becomes free software

June 6th 1999 1.3.95 Beta
   size/speed/memory leak overhaul (no memory leaks)
   XML support for relation loading (for SOLE support)
   JSAPI initial support
   GalaxyCommunicator architecture interface
   ked_mttilt_diphone voice built
   Parser trained on MARSEC (prosodic) brackets rather than syntax
   Unisyn_selection fully integrated
   Unisyn_phonology fully integrated 
   viterbi cart/ngram/wfst base LTS prediction (did improve but BIG)
   viterbi cart/ngram based accent prediction (didn't improve)
   tilt working (again)
   audioin (na_record) for many architectures
   viterbi from Scheme (with cart, ngram, wfst models)

January 26th 1999 1.3.1 Release
   egcs-1.1.1 support
   tobi_rules update (GM)
   replace readline with editline (+ extensions)
   Lots of little bug fixes
   cluster code tidied up
   kal voice
   ked power normalization
   updated lexicons with addenda for US and UK 
   New LTS models for US and UK English
   "Building Voices in Festival" document

August 24th 1998 1.3.0 Release
   UniSun/groupfile optimizations
   Java client support
   Fixed ESPS so both track and wave output works
   Retraining of most modules with new architecture (durations improved)
   rxp, (Richard's XML parser) integrated and Sable XMLified
   Fringe display program for labels and utterances
   Metrical tree synthesis
   A new utterance architecture (Relations and Items)
      utterance save and load work properly now
   Trainable LTS system   
   Lexicon cache system
   Substantial optimization of front end (twice the speed)
   UniSyn, new signal processing and generic waveform synthesis module
   OLS code added
   WFST support for kk rules, regular grammars etc, simple English morphology

November 30th 1997 1.2.4 BETA
   Tilt analysis and Tilt intonation modules added.
   make_utts substantially improved (> 100 times faster)
   text2wave script added
   Pitch synchronous lpc analysis and support
   rab consonant clusters labelled
   New duration tree (wagon stepwise) much smaller if not better
   SCFG grammar and parser (scfg_parse_text added as festival script)
   change config stuff (again)

October 1st 1997 1.2.1 RELEASE
   preliminary support for Visual C++
   Use path-append rather than string-append (in buckets of places)
   Minor bugs fixes throughout the code (end silences are now *always*
     inserted in tts)
   Linux socket bug fixed (get_url didn't work)
   native irix audio support

September 5th 1997 1.2.0 RELEASE
  Proclaim modules and voices
  automatic detection of voices
  Phonset, lexicon, ltsrules listing and printing
  16 bit linear native support for Solaris i386 (sb16)
  Update Festival Tutorial to 1.2.0

Aygust 15th 1997 1.1.99 Beta release
  Win NT (and 95) initial support Cygnus win32 and Visual C++
  100 more pages of documentation
  LPC analysis for voices now ESPS independent
  Spanish el voices tidy up (Borja)
  ToBI by rule implementation
  Confirmed support for gcc-2.7.2, gcc-2.6.3, Linux, FreeBSD, SunOS
    Alpha and SunCC port on Solaris
  reference card added
  return s-expressions in server/client mode
  OGI markup mode added.
  Native support for sun16, linux16 and freebsd (compile time option)
  Changed names of .C files to .cc files for bILL  
  wagon integrated into speech tools (plus docs)
  auto-text-mode-alist for automatic selection of text mode from file name
  Associated token tests added
  Many more tokens dealt with (numbers, money, roman, phone, etc.)
    (analysed databases to see what coverage is like)
  A probablistic chart parser (no significant grammars though)
  RJC's new database/units/join/modify modules taking shape
  Some more examples added to the tutorial (with answers)
  Integrated CSLU changes for OGItoolkit including TCL support
  stml support for phrase types and words inline
  ssml -> stml
  Postlexical rules done in Scheme rather than C++
  Rest of functions to allow any manipulation of utterance from scheme
  New duration models trained for both English and American
  New lexicon (CMU based)
  Consonant cluster support (for kd)
  American diphone set   
  Cluster unit selection algorithm more robust
  Ngram backoff smoothing
  Token pos, for numbers (97.5%) but does poor on phone numbers
  New lexicon with final Rs and r deletion as postlex rule
  Update pos prediction (ts39) and phrase break ngrams (faster to load)
  New ngram format (binary files, and smoothing)
  Vowel reduction module
  Sun CC port
  New string class (rjc) remove dependence on libg++
  Update of course notes and new section on building models from dbs
  Yarowsky homograph disambiguation

Jan 24th 1997 1.1.1 release (first public release)
  a number of configuration and INSTALL documentation bugs fixed
  SSML tidied up and a festival script provided for it.
  Diphones, again, checked and copyright explicitly added

Jan 6th 1997 1.1.0 release
  Roger diphones now default speaker
  A new unit clustering algorithm with acoustic costs and
    optimal coupling
  BSD socket client/server support
  A format function in Scheme (fprintf-like)
  A short course on Speech Synthesis in Festival 
     (with course notes and exercises)
  A programmable form of text modes including externally customizable
     token to word rules.
  Fully programmable intonation module (for ToBI-like theories)
  Backtrace facility in Lisp
  Externally specified Utterance end (for all tts modes) with lookahead
  Roger diphones, first draft

Nov 8th 1.0.0 release
  Substantial bug fixes, stabilization and documentation updates
  Added residual excited LPC synthesizer and removed PSOLA code.
  Made sucs and taylor optional modules, new modules
     can be added without modifying the base code
  MOBY lexicon (not as good as cuvoald but free)
  New diphone grouping software
  A new diphone database module (free from adc)

Sept 30th 0.1 release
  MBROLA support (good example of external module)
  latest news: read out the latest news (from Time Warner, Pathfinder)
  audio spooler
  --language option on command line
  Spanish synthesis
  Letter to sound rules as external system (replacing all the NRL code)
  Welsh synthesis, making the whole system more language independent
  sucs spoke in reasonably way (gsw_450 and f2b dbs)
  document strings for functions (built in and user) and variables
    access from command line and dumped automatically into texinfo
  cleaned up SSML implementation 
  break prediction integrated using viterbi and pos
  sucs module started (selection of units for concatenative synthesis)
  a part-of-speech tagging system (ngram/viterbi based)
  viterbi code added
  fixes in SIOD for running batch and stdin, also sub_prompts added
  saytime example
  Memory leaks fixed, no leaks for tts  
  
July 30th 0.0 release (just for the sake of it)
  a significiant start at documentation (texinfo -> info & html)
  festival scripts using #! on first line
  donovan diphone support 
  can compile (with too many warnings) under g++ 2.7.2
  copyrights on all files
  memory leak checks (only 8 bytes for "unknown" words)
  SSML (and tts file modes)
  cuvoald cmu and beep lexicons
  lexicon compilation
  web page, emacs interface
June 2nd
  Klatt duration module
  syllabification in phones from letter to sound rules
  Linear Regression model for F0 prediction (from ToBI labels)
  CART (wagon) built trees for duration (zscores), phrase boundaries, 
  accent and endtone prediction.
  ffeatures allowing specification of features of an utterance
May  
  integrated Taylor diphone module
  US Naval Research letter to sound rules
  CSTR lexicon

12th April first words "hello"

  start with speech_tools library, scheme-in-one-defun and readline
     and external CSTR diphone synthesizer

7th April 1996 work started