v1.3.0

Highlights

• Expand support to 19 languages with both TN and ITN, plus Hindi–English and Spanish–English code-mixed ITN.
• Add native Korean, German, Armenian, Arabic, French, Portuguese, Vietnamese, Russian, Swedish, Spanish, Hindi, Kinyarwanda, Marathi, Italian, Hebrew, and Hungarian grammars. No NeMo runtime dependency; language-specific ITN graphs remain independent of TN.
• Separate raw-field tagging from semantic verbalization, centralize language capabilities in the CLI, and improve mapping and n-best consistency.
• Unify punctuation spacing and harden number, date, electronic-address, and whitespace boundaries across languages.
• Improve graph construction and test efficiency, including the Chinese TN postprocessor; strengthen configuration-aware FST caching.
• Modernize the Android demo with Kotlin and Compose, and update C++ runtime build support.

Upgrade notes

• Structured tag fields now retain source text. Applications that consume tags directly should check the updated token schemas.
• Install from this release's source to use the new languages. This GitHub release does not itself publish a new PyPI package.

What's Changed

• ⚰️ Remove useless patch & use glog, gflag in utils by @mallocnew in #362
• fix: English TN spurious space after opening punctuation by @pengzhendong in #363
• fix: preserve spaces as token boundaries in English TN tagger by @pengzhendong in #365
• ⬆️ Upgrade CXX standard from 17 to 20 by @mallocnew in #367
• fix: move OOV test cases out of shared test data to fix C++ unit tests by @pengzhendong in #370
• 👷 Ci build for windows by @mallocnew in #371
• 👷 [Runtime] enable unittest on Ci by @mallocnew in #372
• feat(android): build native libs via CMake presets into jniLibs by @mallocnew in #374
• feat(android): migrate app to Kotlin, Compose, and Gradle Kotlin DSL by @mallocnew in #375
• refactor: centralize language pipelines and streamline inference by @pengzhendong in #380
• feat: add Korean TN/ITN with raw-field tagging by @pengzhendong in #381
• feat: add German TN/ITN with raw-field tagging by @pengzhendong in #382
• perf: shrink zh TN postprocessor graph and speed up tests by @pengzhendong in #383
• feat: add Armenian TN/ITN with raw-field tagging by @pengzhendong in #384
• feat: add Arabic TN/ITN with raw-field tagging by @pengzhendong in #385
• feat: add French TN/ITN with raw-field tagging by @pengzhendong in #386
• feat: add Portuguese TN/ITN with raw-field tagging by @pengzhendong in #387
• feat: add Vietnamese TN/ITN with raw-field tagging by @pengzhendong in #388
• feat: add Russian TN/ITN with raw-field tagging by @pengzhendong in #389
• feat: add Swedish TN/ITN with raw-field tagging by @pengzhendong in #390
• feat: add Spanish TN/ITN with raw-field tagging by @pengzhendong in #391
• feat: add Hindi TN/ITN with raw-field tagging by @pengzhendong in #392
• fix: unify punctuation spacing and harden pipelines inherited from NeMo by @pengzhendong in #393
• feat: add Kinyarwanda TN/ITN with raw-field tagging by @pengzhendong in #394
• feat: add Marathi TN/ITN with raw-field tagging by @pengzhendong in #395
• feat: add Hindi-English code-mixed ITN with raw-field tagging by @pengzhendong in #396
• feat: add Italian TN/ITN with raw-field tagging by @pengzhendong in #397
• feat: add Hebrew TN/ITN with raw-field tagging by @pengzhendong in #398
• feat: add Hungarian TN/ITN with raw-field tagging by @pengzhendong in #399
• feat: add Spanish-English code-mixed ITN with raw-field tagging by @pengzhendong in #400

Full Changelog: v1.2.0...v1.3.0