Kana Converter (Hiragana, Katakana, Romaji)

What this tool does

Japanese writes the same sounds in two kana scripts: hiragana for native words and grammatical endings, katakana for loanwords, emphasis and names of foreign origin. Learners, typesetters and anyone cleaning up text copied from a source that used the wrong script need to move between them — and, in a system that cannot display kana, they need roman letters instead.

Method

Hiragana and katakana are parallel blocks in Unicode: hiragana ぁ is U+3041 and its katakana counterpart ァ is U+30A1, a difference of exactly 0x60. Converting between the two scripts is therefore an offset applied to every character in the kana ranges — exact, reversible and free of lookup errors.

Romanisation is a table lookup with two rules layered over it. First, the small kana that combine with a base character are handled as a pair, so きゃ is kya and しょ is sho rather than two separate syllables. Second, the small っ duplicates the consonant that follows, so がっこう becomes gakkou. Katakana input is converted to hiragana before the table is consulted, so one table serves both scripts.

Kana	Romaji	Note
し	shi	Hepburn, not si
ち	chi	Hepburn
つ	tsu	Hepburn
ふ	fu	Hepburn
きゃ	kya	digraph
っ + こ	kko	sokuon doubles the consonant

Example

Input	Katakana	Romaji
ひらがな	ヒラガナ	hiragana
にほんご	ニホンゴ	nihongo
がっこう	ガッコウ	gakkou
おちゃ	オチャ	ocha
とうきょう	トウキョウ	toukyou

Hiragana and katakana are the same sounds

The two kana scripts represent an identical set of syllables. ひ and ヒ are both “hi”; there is no phonetic difference and no meaning carried by the choice of script. What differs is usage:

Script	Used for	Example
Hiragana	Native words, grammatical endings, particles, furigana	たべる (to eat)
Katakana	Loanwords, onomatopoeia, emphasis, scientific names	コーヒー (coffee)

Because the mapping is one-to-one, converting between the scripts never loses information — which is why this tool can do it losslessly in both directions. Converting to romaji is a different matter, because romaji has several competing conventions and some are lossy.

Digraphs and the small kana

Japanese is written in morae, and some morae are spelled with two characters: a consonant kana followed by a small ゃ, ゅ or ょ. These are not separate sounds but a single syllable.

Kana	Romaji	Note
きゃ きゅ きょ	kya kyu kyo	small ゃ/ゅ/ょ modify the preceding consonant
しゃ しゅ しょ	sha shu sho	not “siya” — the digraph is one mora
ちゃ ちゅ ちょ	cha chu cho	same pattern with ち
っ	doubles the next consonant	がっこう → gakkou
ー	lengthens the vowel	コーヒー → kōhī / koohii

The small っ (sokuon) is the one that trips up naive converters. It has no sound of its own; it means “hold the next consonant”, so がっこう is ga-k-ko-u, not gatsukou. A converter that treats it as a normal つ produces visibly wrong romaji.

Why とうきょう becomes “toukyou” and not “tōkyō”

Long vowels are the main source of variation between romanisation systems. Japanese writes a long o either as おう (o + u) or with a macron in some conventions. This tool uses the literal Hepburn-style transliteration: each kana maps to its syllable, so とうきょう → toukyou. That is faithful to the kana and unambiguous to convert back.

The alternatives are トーキョー → tōkyō (macron, common in academic and dictionary usage) or tokyo (dropping the length entirely, as in the English name). All three are “correct” in their own convention; the choice depends on whether you need something a reader can pronounce, something you can convert back, or something that looks like an English word.

Where the conversion is useful

• Looking up a word in a dictionary that indexes by kana or romaji.
• Typing Japanese on a keyboard that produces romaji first — the IME then converts to kana, and the romaji form is what you need to type.
• Sorting and indexing — katakana and hiragana sort in different code-point ranges, so normalising to one script makes an index consistent.
• Reading loanwords aloud: katakana コーヒー → koohii is recognisably “coffee”.
• Text search: a query typed in hiragana will not match text stored in katakana unless both are normalised first.

What is not converted

Kanji are left untouched, because a kanji can have several readings and choosing one requires knowing the word — that is a dictionary lookup, not a transliteration. Punctuation and the prolonged-sound mark are preserved as their own characters. The tool converts the kana scripts and romaji; it does not translate or read kanji.

This is why a page of mixed Japanese text converts partially: the kana come through as romaji and the kanji stay as kanji. That is the honest behaviour — a converter that invented readings for kanji would be guessing, and wrong for any word it had not seen.

Rendering and fonts

Hiragana and katakana need a font that covers the kana block. On a system without one, they render as tofu boxes or fall back to a different style, which can make the two scripts look more different than they are. The conversion itself is independent of the font — it maps code points — but if the output looks wrong, check the font before suspecting the tool.

Frequently asked questions

Go deeper

Longer explanations of the problems this tool solves:

Numeral systems: Unicode digits and Roman numerals

Sources

• Unicode Standard, Hiragana block (U+3040–U+309F) and Katakana block (U+30A0–U+30FF).
• Hepburn romanisation, the conventional romanisation of Japanese for English-language readers.