About
This site is a free reference and practice tool for the three Japanese writing systems: hiragana, katakana and kanji.
How the data is built
Character data — glyphs, romaji readings, stroke counts, row and column positions — is stored as structured data and is the same in every language the site is published in. The explanations around it are written per language rather than machine-translated, because a fluent-sounding but wrong explanation of dakuten is worse than no explanation at all.
Romaji is given in both Hepburn (shi, chi, tsu) and, where they differ, Kunrei-shiki (si, ti, tu), since which one you meet depends on your textbook.
Kanji readings, stroke counts and meanings are derived from KANJIDIC2, published by the Electronic Dictionary Research and Development Group and licensed under CC BY-SA 4.0.
Vocabulary — the written form, reading, part of speech and meaning of every word — comes from JMdict, published by the same group under the same CC BY-SA 4.0 licence. Which JLPT level each word belongs to is not part of JMdict: that comes from the open-anki-jlpt-decks lists (MIT), which trace back to Jonathan Waller’s tanos.co.uk vocabulary lists. Those lists predate the 2010 revision of the JLPT, so a handful of words sit a level away from where a modern textbook would put them.
Stroke-order diagrams use the glyph shapes of Klee One, a handwriting-style typeface by Fontworks Inc. released under the SIL Open Font License 1.1, split into strokes by strokesvg (MIT). The digraphs (きゃ, ファ and the rest) are composed here from their parts. The count, order and direction of every stroke were cross-checked against KanjiVG and AnimCJK.
Pronunciation clips are synthesised with Open JTalk and the MMDAgent Mei voice, licensed under CC BY 3.0. There is one clip per sound rather than per glyph, because あ and ア are the same mora — so the chart plays 125 recordings across its 232 characters, and the kanji pages a further 2,595, one for each distinct reading.
What this site does not do
No account, no paywall, no email capture. Practice progress is stored in your own browser and never leaves the device.
Corrections
Language-teaching content has to be correct, not merely plausible. If something here is wrong, please report it.