Furigana Generator
Paste Japanese text and get the reading printed above every kanji — 漢字 becomes 漢字. That small hiragana line is called furigana (ふりがな), or ruby in typesetting, and it is how Japanese books tell a reader how to pronounce a character they may not know yet. Everything runs in your browser, and you can export it as a Word file where the readings are real, editable Word ruby rather than parentheses glued onto the text.
The analyser runs entirely in your browser. The only thing downloaded is the dictionary itself — your text is never uploaded, so a manuscript or a private letter is safe to paste here.
The grade filter uses MEXT's 学年別漢字配当表, the official list of which kanji are taught in which school year. Set it to grade 4 and 花 or 名前 stay bare while 購入 still gets its reading — which is how a Japanese textbook aimed at a given year actually sets ruby.
No furigana yet
Why a kanji dictionary is not enough
The obvious way to build this is a lookup table: 花 is はな, 火 is ひ, so 花火 must be はなひ. It is not — fireworks are はなび, because Japanese voices the second element of many compounds (a rule called 連濁, rendaku). 大人 is not だいじん but おとな, a reading that belongs to the whole word and cannot be assembled from its two characters at all. And 行った is either いった (“went”) or おこなった (“carried out”) depending entirely on the sentence around it.
So this tool does not use a character table. It runs kuromoji, a Japanese morphological analyser with the MeCab IPADIC dictionary — the same machinery desktop Japanese software uses. It performs a Viterbi search across the whole sentence to decide where one word ends and the next begins, and the reading falls out of the word it picked. That is also why the first run downloads about 17 MB: that is the dictionary, and your browser caches it afterwards.
The readings are split per kanji, not smeared over the word
A word like 食べ物 has the reading たべもの, but printing that as one blob above all three characters is not what a Japanese book does. The reading is aligned to each kanji run — 食(た)べ物(もの) — by matching the kana that already appear in the word (the okurigana) against the reading. When that match is impossible, as with a word whose reading has no per-character basis, the tool falls back to a single reading over the whole word, which is a legitimate style called グループルビ.
Where this is genuinely useful
- Learners reading native material: set the grade filter to your level so only the kanji past it get readings, and you keep practising the ones you know.
- Teachers preparing a handout: export to .docx and the ruby is editable in Word, so you can adjust anything the analyser got wrong.
- Anyone with a Japanese name or address to read aloud — though names are exactly where an analyser is least reliable, and unknown ones are flagged rather than guessed.
अक्सर पूछे जाने वाले प्रश्न
What is furigana, and how does it work?
Furigana is the small hiragana printed above (in vertical text, to the right of) a kanji character to show how it is pronounced. Japanese has roughly 2,136 everyday kanji and thousands more beyond that, and no reliable way to sound one out from its shape - so children's books, newspapers, manga and any text using a rare character print the reading alongside it. In web typography and in Word the same thing is called ruby text. It is purely a reading aid: it adds no meaning, and a fluent reader simply ignores it.
Is this furigana generator free, and does my text get uploaded?
It is free with no signup and no limit on how many times you use it, and your text is never uploaded. The morphological analyser and its dictionary are downloaded to your browser and everything runs locally on your own device - the same is true of the Word and PDF exports, which are assembled in the browser. That means a manuscript, a client letter or an unpublished translation is safe to paste in.
How accurate is it? Where does it get readings wrong?
For ordinary prose it is very good, because it picks the word before it picks the reading - which is what gets 花火 as はなび rather than はなひ, 大人 as おとな, and 行った read correctly from the surrounding grammar. It is weakest on three things. Personal and place names, where a single spelling can have several legitimate readings and the dictionary may have none at all; those are left bare and marked rather than guessed. Deliberately unusual readings an author invented for effect. And numbers with counters, where the counter changes the reading of the number. Treat the output as a very good first pass that a Japanese reader should check, not as an authority.
What is the grade filter, and which level should I pick?
Japan's Ministry of Education publishes the 学年別漢字配当表, which assigns 1,026 kanji to specific elementary school years and treats the remaining jōyō kanji as junior-high-and-above. The filter uses that list: set it to grade 4 and every kanji taught in grades 1 to 3 is left bare while harder ones still get readings. Pick the grade you can already read comfortably. Publishers do exactly this - a book for nine-year-olds sets ruby only on kanji above that year's list, which is why the same word can appear with furigana in one book and without it in another.
Does the Word export keep the furigana above the kanji, or turn it into parentheses?
Above the kanji, as real ruby. The .docx contains genuine WordprocessingML ruby runs - the exact structure Word produces when you use its own ルビ button - so the reading sits above the base text, stays attached to it when you edit the sentence, and can be resized or removed with Word's ruby dialogue. Most converters flatten furigana into 漢字（かんじ）, which is fine for a plain-text note and useless for anything you intend to print or hand out. There is a copy-as-parentheses button here too if that is what you actually want.
Can I get katakana readings instead of hiragana?
Yes, there is a switch for it. Hiragana is standard for furigana in books and is what you want almost always. Katakana readings are what Japanese official forms ask for in the フリガナ field next to your name - the box where you write how your name is pronounced. If you are filling in a Japanese form rather than reading a book, that is the setting you want.
Why does the PDF sometimes say characters were dropped?
A PDF has to carry its own font for Japanese, since PDF's built-in fonts cannot represent a single kana. The font embedded here is Sawarabi Mincho under the SIL Open Font License, chosen because it is 3.5 MB rather than the 7.7 MB of the usual alternative - and it covers JIS X 0208 level 1, which includes every jōyō kanji but not the rarer level-2 characters. Rather than emit a PDF full of blank boxes, the tool checks coverage first and tells you exactly which characters it could not draw. The Word and HTML exports are unaffected, because they use whatever font the reader has installed.
Can I use the output on a website?
Yes - the HTML download is standard <ruby> markup with <rt> for the reading and <rp> fallback parentheses, which every current browser renders natively with no JavaScript or library. The <rp> tags matter: anything that strips tags, including most paste targets, falls back to 漢字（かんじ）rather than running the reading into the text. There is also a vertical (縦書き) HTML export that sets writing-mode: vertical-rl, which is how Japanese prose is traditionally laid out.