What furigana is for
Furigana are the small readings printed above kanji, familiar from children’s books and manga for young readers. For a learner they provide a temporary scaffold: with them you can read a text whose kanji you have not learned; without them you can find out which words you recognise. They explain pronunciation, rather than the meaning of a word. A dictionary or translation may still be useful.
How to use it honestly
Read a short passage with furigana once for meaning, then read it again with the readings hidden. Notice which words you can recognise on the second pass. If you always look at the kana first, try covering just one familiar word’s reading. There is no need to remove every reading at once. The checkbox above keeps the text in place while hiding its furigana, so you can compare both versions without losing your place.
Where readings go wrong
Japanese kanji have different readings depending on context and the words around them. A character-by-character converter can produce the wrong answer: the 生 in 生まれる is not read as せい. This tool uses kuromoji with the IPADIC dictionary to identify words before adding readings. It handles many ordinary words, but it can still miss names, coined terms and unusual compounds. A familiar-looking reading is not a guarantee. Check names and anything you plan to publish against a reliable dictionary or with a Japanese reader.
Okurigana belong outside the reading
Okurigana are the hiragana written after a kanji stem, like the べる in 食べる. They belong outside the reading. 食べる becomes 食[た]べる: the べる is already readable and does not need a second reading above it. This converter aligns the kana in each word with its dictionary reading. Some compounds have a reading that cannot sensibly be assigned character by character. 今日, read きょう, stays together as one reading group. Where a more detailed alignment would be uncertain, the tool keeps the word together rather than inventing individual kanji readings.
Choosing an output format
HTML ruby is useful on a webpage you control. The preview shows how that markup reads, and the HTML field gives you the escaped source to copy. Bracket notation, such as 漢字[かんじ], travels more reliably through spreadsheets, flashcards and messages. Plain kana replaces words that have readings while keeping existing kana, punctuation and Latin text. Unknown words are left unchanged and listed in the result; they are not silently guessed.
Longer passages and privacy
You can convert up to 10,000 characters at once. The reading dictionary (about 17 MB) downloads when you first press the button, so the first conversion may take longer than later ones; on mobile data, wait for Wi-Fi. Conversion happens in a background worker in your browser. Your pasted passage is not sent to a conversion server. For a long text, work in paragraphs so that you can check unfamiliar readings as you go, especially dialogue, proper names and specialist vocabulary.
Reading engine: kuromoji.js, Apache 2.0. Dictionary: IPADIC, with the dictionary notices and engine license distributed alongside it. Whole-word corrections such as 送り仮名 and 一人 come from JMdict, EDRDG, under CC BY-SA 4.0; the selected entries and changes are shared under that license.