Pass HanziToAnki a file via command-line or web interface, and generate flashcards for Anki/Pleco/Memrise.

You can generate the flashcards from any Chinese input - a news article you're studying, song lyrics, or even your exported WeChat logs!

HanziToAnki supports two modes: API Server (web interface) and CLI Tool (command-line). The same JAR/Gradle build supports both modes automatically.

Run with JAR:

java -jar build/libs/HanziToAnki-1.0.0.jar

Run with Gradle:

./gradlew runApi

Then open http://localhost:8080/ in your browser to upload files and generate decks.

Run with JAR:

java -jar build/libs/HanziToAnki-1.0.0.jar input.txt
java -jar build/libs/HanziToAnki-1.0.0.jar input.txt -f ANKI -o output.anki

Run with Gradle:

./gradlew runCli -Pargs='input.txt'
./gradlew runCli -Pargs='input.txt -f ANKI -o output.anki'

Command-line options:

• -w --word-list Read from an input file containing a list of words, separated by line breaks. Without this flag, individual characters are extracted
• -s --single-characters Extract only single characters from the file
• -hsk <hsk level> Remove any words in any HSK levels up to and including the given one
• -t --strategy <strategy> Specify the word finding strategy. Options:
  • 6 - ANSJ_SEGMENTATION (default): Intelligent word segmentation using ANSJ (Jieba-like)
  • 0 - TRI_BI_MONOGRAMS_USE_ALL_CHARS_BIGRAM_OVERLAP: 3-char, 2-char, 1-char combinations
  • 1 - TRI_BI_MONOGRAMS_USE_ALL_CHARS: Only 3-char combinations
  • 2 - BIGRAM_AND_MONOGRAM_ONLY_NO_OVERLAP: 2-char and 1-char, no overlap
  • 3 - BIGRAM_AND_MONOGRAM_ONLY_OVERLAP: 2-char and 1-char with overlap
  • 4 - SINGLE_CHAR_ONLY: Single characters only
  • 5 - ALL_COMBINATIONS: All possible n-gram combinations
• -o <output filename> Override the default output file name
• -f --format <output format> Override the default output file name (ANKI, PLECO, MEMRISE)
• -c --char-type <char type> Specify the type of character (TRAD, SIMP or SIMP_AND_TRAD)

Features:

• Both simplified and traditional hanzi can be provided
• English definitions provided by CC-CEDICT
• Pinyin has accented characters (nĭ hăo instead of ni3 hao3)
• Pinyin is coloured by HTML markup
• Can ignore vocabulary below specified HSK level (reduces card count & saves time)
• Fully Open-Source, so you can contribute features, create issue tickets on GitHub, or even help fix those issues!

Please feel free to make suggestions, open/comment on issues, or share code!

Feel free to fork, create branches, and raise PRs.

If you get an error about "Invalid source release", check that echo $JAVA_HOME points to your JDK. We recommend sdkman for setting up Java.

The project uses Java 21 and Gradle 8.9.

Build the dual-mode JAR (supports both API and CLI):

./gradlew build

The JAR will be at build/libs/HanziToAnki-1.0.0.jar

See Running HanziToAnki above for all running options.

Quick reference:

• API server: ./gradlew runApi or java -jar build/libs/HanziToAnki-1.0.0.jar
• CLI tool: ./gradlew runCli -Pargs='input.txt' or java -jar build/libs/HanziToAnki-1.0.0.jar input.txt

Run all tests:

./gradlew test

The project includes:

• 40+ unit tests for core functionality
• 15+ integration tests for output validation with modern Chinese fixtures
• Tests for all 7 word segmentation strategies

You can also run tests with IntelliJ or other IDEs.

This project uses a modified version of the CEDICT Chinese dictionary, which can be found here: https://cc-cedict.org/wiki/