Releases: RHVoice/RHVoice
Releases · RHVoice/RHVoice
1.8.0
What's Changed
- Fix NVDA driver availability without installed voices by @kvark128 in #443
- add Polish voice by @grzezlo in #445
- Warn about the need to install voices in the NVDA addon by @kvark128 in #446
- A build on Linux shouldn't fail without Speech Dispatcher. by @Olga-Yakovleva in #454
- mgc with world by @Olga-Yakovleva in #459
- voice-building-utils correction and convert-audio to py3 migration. by @grzezlo in #460
- Partially parallel context clustering. by @Olga-Yakovleva in #465
- Core: handle binary feature flags in the syllabification output. by @Olga-Yakovleva in #466
- Voice training: support additional iterations of tree context clustering. by @Olga-Yakovleva in #467
- Update Ukrainian dictionary by @VovaMobile in #468
- SSML: support the phoneme element, only native phonemes are supported. by @Olga-Yakovleva in #469
- Voice-building-utils: save the trim positions in the original recordings. by @Olga-Yakovleva in #472
- New male polish voice: Natan by @zstanecic in #473
- NVDA addon: Move Rate Boost setting to the conventional position. by @TheQuinbox in #490
- Segmentation, experimental: use vad to initialize the silence model. by @Olga-Yakovleva in #491
- Voice-building-utils: the export-voice command calculates and saves the key parameter. by @Olga-Yakovleva in #500
- Core: sentence level tokenization support. by @louderpages in #501
- Native data package manager, Android-only for now. by @Olga-Yakovleva in #452
- Add script to convert text files to SSML format by @cyrmax in #504
- NVDA addon: Make RHVoice compatible with NVDA 2022.1 by @zstanecic in #507
- Macedonian Language Updates by @louderpages in #508
- Fixe segmentation fault (core dumped) on Georgian language by @Olga-Yakovleva in #516
- Improved support for the reuse of msi package files in SAPI voice installers. by @Olga-Yakovleva in #523
- cmake: Support new speech-dispatcher modules location by @living180 in #535
- RHVoice-test: don't ignore the quality setting in the configuration file. by @Olga-Yakovleva in #538
- NVDA addon: correct the project website url. by @Olga-Yakovleva in #539
New Contributors
- @savoptik made their first contribution in #429
- @kvark128 made their first contribution in #443
- @TheQuinbox made their first contribution in #490
- @louderpages made their first contribution in #501
- @cyrmax made their first contribution in #504
- @gerazov made their first contribution in #534
- @living180 made their first contribution in #535
Full Changelog: 1.6.0...1.8.0
1.6.0
- Albanian language
- Voices: add tatiana (#386)
- Voices: add Mikhail
- Voices: add Marianna (#400)
- Voices: add lyubov (#412)
- Voices: add Vitaliy
- Voices: add Albanian voice, Hana, North Macedonian accent
- Update emoji data to version 40
- Android: Clean up and update dependencies.
- Android: fix the version display if there are no voices installed.
- Android: users can now select a dictionary file on their device and import it into the app.
- Android: don't create an initial version of the config file.
- Android: the option to import a config file.
- Android: target Android 11.
- Android: fix a crash.
- Android: correct some translations.
- Core: a language developer may set custom flags on segments in post-g2p and use them in their hts labels.
- Core: g2p.case setting.
- Core: word_break feature.
- Core: more hts features.
- Core: data-only languages.
- Core: data-only languages get their list of graphemes from a file.
- Core: voices may provide equalization filters.
- Use a cascade of biquad filters.
- Save the coefficients in 24000/eq.txt.
- First line is a header with the version number. In case I decide to change the format.
- Then coefficients separated by whitespace.
- For each filter: b0, b1, b2, a0, a1, a2.
- Core: enable the limitter if equalization is used by the voice.
- Core: general support for a simple "accented" decision tree.
- Core: Failing to load the eq shouldn't cause the voice loading to fail.
- Core: support comments in eq.txt.
- Core: if the name of a phone starts with /, remove the / after syllabification.
- Thvoice-transcribe-sentences: the -v option now also prints out gpos if not content.
- Brazilian Portuguese: fix the pronunciation of the word backup.
- Brazilian Portuguese: pronunciation of the ordinal indicator characters.
- update Ukrainian dictionary
- honor 'SOURCE_DATE_EPOCH' environment variable if it's set.
this helps rhvoice build reproducibility.
version 1.4.2
RELEASE 1.4.2