21.03A1

This release brings in all changes from the parent repository, mainly consisting of localization updates.

21.03-x0_personal

This release fixes compatibility with the NVDA 2021.1 alpha snapshots by updating to use the proper imports and references, since 2021.1 has deprecated some older versions of them we were using.

21.01-x0_personal

Important: Version compatibility

This version drops support for NVDA 2019.2.1 and prior.
Changes in this release:

• Dictionary handles are now cleared on exit. This should fix loading alternate language dictionaries in certain cases, as well as making it reload dictionaries whenever reloading the synth. Thus you don't have to restart NVDA anymore to load dictionary updates.
• Fixed synth switching to the wrong language when loaded.
• Fixed some loopholes created when anticrash was refactored.
• Fixed language identification.
• Updated anticrash
• Added a checkbox to toggle always sending speech parameters. When enabled the rate bug should not occur, but pauses are overly dramatic when using IBM ViaVoice binaries.
• Added conditional support for IBM ViaVoice binaries, with their own fixes.
  Many thanks go to @Mohamed00 for the pull requests integral to this release, and for performing testing and integration for the viavoice binaries.

20.10-x0_personal

This release includes the following improvements:

• Added a workaround for the date parser false positives, fixing some rather strange wordings for phrases involving numbers and what might be a partial month name that is part of a bigger word.
• Removes fix expressions for '.' and '@' in all languages. NVDA does not pass stray dots by default, and @ is never sent.
  • If you prefer your dots sent through in the symbols dictionary but do not wish ECI to spell anything on the right side, add this to your voice dictionary.
  • Pattern: ([a-zA-Z0-9])\.([a-z])
  • Replacement: \1 dot \2
  • Type: Regular expression.
• Fixed the Asian language locale identification to NVDA. Trust voice's language features should now work on these languages.
• Other improvements.

20.09-x0_personal

Changes in this release

• Further updated anticrash expressions, adding one for NVDA splitting cammel-case names up when said cammel-case name was in a roots dictionary.
• Makes pause shortening expression a little more robust, to work in more cases when encountering chained punctuation.
• Adds proper support for loading different user dictionaries for different languages, at the time of switching to that language.
• If libraries are bundled inside the add-on, config library path is ignored to prefer the internal bundle.
• Added pause adjustment by rate logic from eloquence_threshold, used when encountering a pause command in a speech sequence from NVDA.

20.08.1-x0_personal

This release fixes an anomaly presented when one had shorten pauses enabled and encountered text that required special handling by ECI, from user dictionaries or internal filters alike. If you use the shorten pauses feature, it is strongly advised to upgrade to this release.

20.08-x0_personal

This release updates pause shortening logic to be more robust, catching consecutive punctuation, allowing whitespace on the left side of the punctuation to be shortened, and adding en dash and em dash to characters considered punctuation to be shortened. Furthermore, now supports loading of pronunciation dictionaries other than US English depending on language selected at initialization time of the synthesizer.

IMPORTANT NOTICE

As of this release, IBMTTS will no longer fall back to root.dic/main.dic/abbr.dic if enuroot.dic/enumain.dic/enuabbr.dic are not present. To continue using existing files with these names, rename them as needed, prefixing them with the three letter language code.

Also of note

Regarding pronunciation dictionaries, please see the community dictionary repository for expansive pronunciation dictionaries, and a place to submit your own entries to any language of your choosing.

20.07.2-x0_personal

This release updates anticrash expressions again.

20.07.1-x0_personal

This release updates the tz crash expression, sets the maximum eci rate without rate boost back to 150 instead of 156, adds some German crash expressions to those filtered, removes apostrophe expression false positive on "shouldn't've," and fixes type errors encountered with some regexes which caused the string to not be spoken at all.

20.07-x0_personal

This release updates built-in anticrash expressions, introduces toggles for using the abbreviation dictionary, ECI phrase prediction and pause shortening.
The NVDA add-on hosted here includes no binaries. Given the existence of the Community Dictionary Repository, it is best practice to store your IBMTTS binaries outside the add-on, such as the functionality that allows one to install it into a separate binaries-only add-on. This would make updating your root and other dictionaries easier and updates to IBMTTS non-destructive in this regard.