v0.24.1

What's Changed

Fixes

• Fix Espeak issues on Windows by @eginhard in #34
• Fix XTTS streaming by @eginhard in #33

Full Changelog: v0.24.0...v0.24.1

v0.24.0

What's Changed

Features

• Move from setup.py to pyproject.toml, simplify requirements by @eginhard in #19
  This removes language-specific G2P dependencies for Bangla, Korean and Mandarin Chinese from the default install. They must now be installed via optional extras, as is already the case for Japanese: https://github.com/idiap/coqui-ai-TTS?tab=readme-ov-file#optional-dependencies

Fixes

• fix(espeak_wrapper): capture stderr separately by @eginhard in #17
• fix(bark): add missing argument for load_voice() by @eginhard in #22
• fix(server): ensure logging output gets actually shown by @eginhard in #26

Full Changelog: v0.23.1...v0.24.0

v0.23.1

What's Changed

Features

• Python 3.12 support by @eginhard in #11

Fixes

• fix(LanguageManager): allow initialisation from config with language ids file by @eginhard in #9

Full Changelog: v0.23.0...v0.23.1

v0.23.0

What's Changed

Features

• Use Python logging instead of print() by @eginhard in #3, #5
• feat(xtts): support Hindi for sentence-splitting and fine-tuning by @eginhard in #4

Full Changelog: v0.22.1...v0.23.0

v0.22.1

What's Changed

Fixes

• Bug fix in MP3 and FLAC compute length on TTSDataset by @Edresson in coqui-ai#3092
• Fix get_git_branch() in generic_utils.py by Nick Potafiy in coqui-ai#3561
• fix(api): handle missing attribute in is_multi_lingual by @eginhard in eginhard#4
• Fix TTS().list_models() by @greerviau in eginhard#5
• Fix bark model url by @ipeevski in eginhard#6
• fix(xtts_manager): name_to_id() should return dict by @eginhard in eginhard#8
• fix chinese pinyin phonemes by @JaysonAlbert in eginhard#10
• Fix synthesis with fairseq models by @akgupta1337 in eginhard#11
• build: update numpy and librosa by @eginhard in eginhard#14
• docs(tts.models.vits): clarify use of discriminator/generator by @eginhard in eginhard#19
• fix: torch.stft will soon require return_complex=True by @eginhard in eginhard#20

Refactors, dependency updates

• Remove pandas dependency by @eginhard in eginhard#3
• build: remove unused dependencies by @eginhard in eginhard#12
• build: make dependencies for server optional by @eginhard in eginhard#15
• Remove duplicate Trainer code by @eginhard in eginhard#18
• refactor(dataset): get audio length with torchaudio by @eginhard in eginhard#21
• refactor(bin.find_unique_chars): use existing function by @eginhard in eginhard#22

Dev improvements

• Fix CI (save best model after 0 steps in tests) by @eginhard in coqui-ai#3423
• CI: remove unused check_skip steps by @akx in eginhard#2
• Lint overhaul (pylint to ruff) by @akx in eginhard#1
• Pin black for consistent outputs by @eginhard in eginhard#7
• test(vocoder): disable wavegrad training test in CI by @eginhard in eginhard#9
• ci: combine coverage reports and merge test workflows by @eginhard in eginhard#16
• Update readme by @eginhard in eginhard#17
• Update repository links, package names, release script by @eginhard in eginhard#24

Full Changelog: v0.22.0...v0.22.1