These open dictionaries and fonts make Wing Font's dialect annotations possible. Thank you to everyone who shared their work.
Wing Font is a personal open-source project. The author is not a trained linguist. Every romanization table shipped with the pre-built fonts originates from one of the upstream open-source sources credited below — but those sources are heterogeneous in format, vocabulary, and tone-mark convention, so the data has been normalised, deduplicated, and disambiguated using a mix of hand editing, regex, and AI-assisted tooling (LLMs for cleanup; programmatic transformation for tone-mark and segmentation conversions; generated fallback readings where coverage gaps existed).
This means errors are likely — wrong default readings for polyphonic characters, missing dialect-specific variants, occasional mis-segmented multi-character entries, edge cases around tone sandhi.
Native speakers, linguists, and curious readers who spot a mistake are warmly invited to file a GitHub issue or join the Telegram group — fixes ship in the next CI build.
The standard (優勢腔) readings and the nine accent (腔) variants come from the MOE 臺灣台語常用詞辭典 (sutian / kautian.ods, CC BY-ND 3.0 TW). Tâi-lô and POJ are taken straight from the dictionary; TLPA and 閩拼 are derived from Tâi-lô. Multi-character words carry one syllable per character so they drive 多音字 disambiguation.
MOE 臺灣台語常用詞辭典 (sutian / kautian.ods) — the authoritative source for standard character readings and the per-accent 語音差異 table. Licence: CC BY-ND 3.0 Taiwan.
ChhoeTaigi open database — the faithful conversion of the MOE dictionary whose KipUnicode (Tâi-lô) / PojUnicode (POJ) columns Wing Font actually consumes.
Transcription data for 方音符號 (Taiwanese Phonetic Symbols) and 台灣語假名 (Taiwanese kana) (Apache-2.0); readings sourced from the MOE Taiwanese dictionary.
The Teochew character and word readings (Geng'dang Pêng'im) come from the open learn-teochew project, then converted into Peng'im / Tie-lo / Pe̍h-ūe-jī and other schemes with the parsetc parser.
The Teochew character / word reading dictionary (teochew_scrape.json, Geng'dang Pêng'im) — also the source of the home-page sample line 「家己个歌，家己唱；家己个字，家己揀。」.
Teochew romanization parser and converter — turns the Peng'im readings into Tie-lo, Pe̍h-ūe-jī, Dieghv, Gaginang and Sinwenz.
Every TTF Wing Font ships with is licensed under the SIL Open Font License 1.1 (OFL-1.1). The canonical OFL text and each font's copyright statement + Reserved Font Name are bundled together in the wing-font-hub repo's LICENSES/ folder. The list below groups fonts by role.
Traditional-Chinese CJK base font, variable weight. SIL Open Font License 1.1.
Simplified-Chinese CJK base font, variable weight — pairs with the Mandarin pinyin showcase. SIL Open Font License 1.1.
Hong-Kong-locale CJK base font, variable weight — default base for the Cantonese showcase fonts. Glyph forms follow HK conventions (different stroke shapes for chars like 字 / 為 / 起 / 緣 / 緊 vs the Taiwan-locale Noto Sans TC). SIL Open Font License 1.1.
Annotation font for the Cantonese-katakana mapping (variable weight). SIL Open Font License 1.1.
Annotation font for the Cantonese-Hangul mapping (variable weight). SIL Open Font License 1.1.
Traditional-Chinese serif (Sung) base font; ships in Regular and Italic. SIL Open Font License 1.1.
Traditional-Chinese sans-serif (Hei) base font; ships in Regular and Bold. SIL Open Font License 1.1.
Pan-CJK serif, Adobe's Source Han Serif — paired with the M+ 1m annotation in one of the Mandarin showcase combinations. SIL Open Font License 1.1.
Simplified-Chinese handwriting-style base font (lxgw's Xiaolai SC) — paired with M+ Rounded 1c for a softer Mandarin pinyin showcase. SIL Open Font License 1.1.
Annotation font — justfont's open-source Huninn (jf-openhuninn). Carries every Tâi-lô / POJ combining tone mark plus the nasal ⁿ and the o͘ dot. SIL Open Font License 1.1.
Latin annotation font used for the Cantonese romanizations (LSHK, Yale, Lau, Guangdong, Chishima). SIL Open Font License 1.1.
Japanese fixed-width annotation font, paired with Source Han Serif for one of the Mandarin showcase combinations. SIL Open Font License 1.1.
Japanese rounded annotation font, paired with Xiaolai SC for the handwriting-style Mandarin showcase. SIL Open Font License 1.1.
Annotation font for the Cantonese-Thai mapping (variable weight), designed by Cadson Demak / IT Foundry. SIL Open Font License 1.1. "Google Sans" is a trademark of Google LLC.
Jyutping (LSHK) reading data.
CUHK's 粵語審音配詞字庫 (Chinese Character Database with Cantonese readings).
Cantonese Romanization Converter.
The Mainland 普通話 reading data (mandarin-cn) is built entirely from mozillazg's MIT-licensed datasets — taken from their permissive upstream, not from any OFL-licensed font project — with each character's default reading following the Unicode Unihan database's kMandarin field. Singapore and Malaysia officially adopt the same 普通話 standard, so they share mandarin-cn. The Taiwan 國語 variant (mandarin-tw) diverges: 753 single-character default readings are re-derived from the Taiwan Ministry of Education's 《國語辭典》 (e.g. 期 qí, 危 wéi, 突 tú), sourced from g0v's moedict-data. The technique of baking pinyin into the glyphs with contextual homograph disambiguation was pioneered by Mengshen-pinyin-font (OFL-1.1); Wing Font reuses none of its data files but gratefully acknowledges it as inspiration.
Per-character 普通話 readings (MIT; derived from Unihan).
Word/phrase readings that drive contextual homograph disambiguation (MIT).
Unicode Han Database (Unihan) — upstream reading data.
Taiwan MOE 《國語辭典》 readings used for the 國語 variant (mandarin-tw), via g0v's moedict-data (CC BY-ND 3.0 TW).
Each dataset and font is used under its own licence — check the upstream source before redistributing.