version 23.4.1 and the last changes.

Changed eciSynthMode from the default 0 (Sentence) to 1 (Manual): Synthesis and input clearing is controlled by commands only.
Changed the name of callback to eciCallback, and the decorator was placed just before the function.

addon/synthDrivers/_ibmeci.py

@@ -165,8 +165,9 @@ def run(self):
 		msg = wintypes.MSG()
 		user32.PeekMessageA(byref(msg), None, 0x400, 0x400, 0)
 		(dll, handle) = eciNew()
-		dll.eciRegisterCallback(handle, callback, None)
+		dll.eciRegisterCallback(handle, eciCallback, None)
 		dll.eciSetOutputBuffer(handle, samples, pointer(buffer))
+		dll.eciSetParam(handle, ECIParam.eciSynthMode, 1)
 		dll.eciSetParam(handle, ECIParam.eciInputType, 1)
 		params[ECIParam.eciLanguageDialect] = dll.eciGetParam(handle, ECIParam.eciLanguageDialect)
 		# loading of fallback root.dic/main.dic/abbr.dic officially removed as of 20.08-x0_personal, to make room for other languages' dictionaries.
@@ -351,9 +352,9 @@ def playStream():
 	sendIndexes()
 
 endStringReached = False
-Callback = WINFUNCTYPE(c_int, c_int, c_int, c_int, c_void_p)
-@Callback
-def callback (h, ms, lp, dt):
+
+@WINFUNCTYPE(c_int, c_int, c_int, c_int, c_void_p)
+def eciCallback (h, ms, lp, dt):
 	global audioStream, speaking, END_STRING_MARK, endMarkersCount, indexes, endStringReached
 	if speaking and ms == ECIMessage.eciWaveformBuffer:
 		audioStream.write(string_at(buffer, lp*2))

changelog.md

@@ -1,17 +1,9 @@
-# version v23.02.1
-
-* Fixed some bugs in some cases when loading IBMTTS libraries. The way of loading the libraries has been changed to minimize the possibility of errors.
-* If the driver can't load the library, an error entry will be shown in the log.
-* Changed the way to determine the library being used. Now dll resources (ProductName) is used to determine it.
-* Added the version where a  documentation change was introduced (for documentation translators).
-* Added fixes for french.
-* Updated code to use language constants.
-* Improved the conditionals to determine the language to be used in process text.
-* Restored the 22 kHz sample rate option. Voices that use it have surfaced recently.
-* Removed outdated version of Spanish ª crash fix from spanish_ibm_fixes.
-* Updated spanish documentation.
-* Updated locale strings and documentation for Brazilian and Portugal Portuguese.
-* Updated french documentation.
-* Updated german interface and documentation.
-* Updated translations for italian language.
-* Updated english documentation.
+# version 23.4.1
+Changed eciSynthMode from the default 0 (Sentence) to 1 (Manual): Synthesis and input clearing is controlled by commands only.
+Tried to work around situations where IBMTTS would say 'comma hundred' or other weird things in English if numbers are separated by commas
+Some fixes in english_ibm_fixes should've actually applied globally. Moved the punctuation fixes formerly part of english_ibm_fixes to a new global_ibm_fixes section to improve the experience for non-English languages
+Fixed short pauses pronouncing the right parenthesis if IBMTTS was used. While this change will make the short pauses expression slightly less affective, it will still work in general
+Updated German anticrash further to catch more cases. #88
+More space removal changes. The expression will take into account numbers as well as letters in the negative lookahead, and added the question mark to its captured punctuation
+Improved the IBMTTS space removal expression so it doesn't catch false positives such as it's a .mp3 file
+added a log message if the selected library is not correct. It shows the exception that raised the failure.