Cantonese Writer
Create Anki flashcard decks for learning Cantonese — stroke-order writing practice, Jyutping, Traditional characters, and native TTS audio.
A Cantonese adaptation of Anki-xiehanzi.
Live app: https://joistad.github.io/cantonese-writer/
What it does
- Look up Cantonese vocabulary from a built-in dictionary (CC-CEDICT + CantoJpMin)
- See Traditional Chinese characters with Jyutping romanisation and English definitions
- Export an Anki
.apkg deck using the Cantonese Writer note type
- Cards include a HanziWriter stroke-order practice field — draw the character on screen and get it graded stroke by stroke
- Optional Cantonese audio via Google Cloud TTS (Hong Kong yue-HK voice — genuine Cantonese, not Mandarin)
- Tone colours follow the 6-tone Cantonese system
How to use
- Open the live app
- Configure your deck (card layout, audio settings)
- Add the characters you want to learn
- Click Generate Deck — downloads a
.apkg file
- Import the file into Anki
Card fields
| Field | Content |
|---|
| Simplified | Traditional Cantonese character(s) |
| Traditional | Same as Simplified |
| Pinyin | Jyutping romanisation |
| Zhuyin | (empty) |
| Definitions | English definitions with tone-coloured Jyutping |
| Audio | [sound:yue-word.mp3] if audio was generated |
Notes on character coverage
HanziWriter stroke data comes from hanzi-writer-data, which covers ~9,500 characters from Standard Chinese. Characters exclusive to written Cantonese are mostly absent from it.
For the most common of these, Cantonese Writer ships generated stroke data in stroke-data/. Each file is composed from existing hanzi-writer-data strokes (e.g. 喺 = the 口 of 唔 + a scaled 係); nothing is hand-drawn. The app and the card template use the official dataset first and fall back to these files. See stroke-data/README.md for the method.
Covered by generated data: 167 characters, including the most common Cantonese ones (佢 喺 咗 啱 冇 啲 嚟 咁 哋 㗎 搵 諗 嗰 噉 喎 攞 嗱 黐 啩 抌 孖 餸 冧 踎 揼 劏 𨋢 嚡 …). The full ranked list is in stroke-data/targets.tsv, and what is and isn't covered is in stroke-data/README.md.
嘅, 唔 and 係 were listed as unsupported in earlier versions of this README, but they have always been in the official dataset.
Any other character without stroke data is skipped when you add words, and the skipped words are listed so you know.
Existing decks: the fallback lives in the card template, which Anki stores with the note type when you first import a deck. Decks imported before this change keep the old template. To get the fallback, either generate and import a new deck, or copy the updated card template into Anki (Tools → Manage Note Types → Cantonese-xiehanzi → Cards).
Tech stack
- Vanilla JS / HTML / CSS — no build step, runs entirely in the browser
- genanki-js for
.apkg generation
- HanziWriter for stroke-order animation and quiz
- CantoJpMin for Jyutping lookup
- CC-CEDICT as the base dictionary
- Google Cloud TTS (yue-HK-Standard-A) for Cantonese audio
- Anki Persistence embedded in card templates
Credits
This project is a Cantonese adaptation of Anki-xiehanzi by @krmanik. Card templates, note type structure, and core .apkg generation logic are derived from that project.
Jyutping data from CantoJpMin by @hanleyweng.
License
Card template data is derived from Xiehanzi (MIT) and hanzi-writer-data (Arphic Public License).
Generated stroke data in stroke-data/ is derived from hanzi-writer-data and distributed under the same terms (Arphic Public License / LGPL-3.0).
Dictionary data from CC-CEDICT (Creative Commons Attribution-Share Alike 3.0).
