Convert hiragana and katakana to romaji in every system in current use — the official 2025 standard, Hepburn, passport, Kunrei-shiki and Nihon-shiki — side by side.

RT-TXT-060 · Text Tools

Romaji Converter

Kana only — half-width katakana works too. Kanji cannot be converted: a character's reading depends on the word it is in, so type the reading.
Long vowels (おう, うう, ああ…)
Try:
The same kana in each romanization system
Official (2025) 令和7年内閣告示第4号, 22 Dec 2025 — macron form …
Official (2025), vowels doubled the same notice's second long-vowel form …
Hepburn (modified, ALA-LC) dictionaries and library catalogues …
Passport Ministry of Foreign Affairs Hepburn table …
Kunrei-shiki 1954 notice 第1表 · ISO 3602 — abolished 22 Dec 2025 …
Nihon-shiki the 1954 notice's 第2表 spellings · ISO 3602 Strict …

A marked row spells this input differently from the official 2025 standard.

📅 Research current as of 27 Sep 2026 · Sources: 令和7年内閣告示第4号「ローマ字のつづり方」(22 December 2025), which abolished 昭和29年内閣告示第1号 (9 December 1954); its 対照表 for the 1954 第1表 (Kunrei-shiki) and 第2表 (Nihon-shiki) spellings; the Ministry of Foreign Affairs passport Hepburn table (外務省ヘボン式ローマ字綴方表); loanword combinations as in the 1991 notice on writing loanwords (外来語の表記).
Standards, published formulas, and reference data are revised over time — check against a current authoritative source before relying on these figures.

Everything runs in your browser — nothing you type is sent anywhere.

Advertisement

How to use the romaji converter

1
Type or paste kana

Hiragana, katakana or both, including half-width katakana. If you only have kanji, look up the reading first — the tool will not guess one.

2
Read the six rows

Each row follows one published system. Rows marked in the margin spell your input differently from the official 2025 standard, so you can see at a glance where the systems part company.

3
Settle the long vowels

By default おう, うう and similar pairs are read as one long vowel, which is right for most words. If the note under the table flags a pair that belongs to two different sounds — 思う omou is the usual case — switch to two separate vowels.

4
Copy the row you need

Click a spelling to copy it. Use the passport row for a passport or visa form, the official row for new public signage and documents, and ALA-LC for library records.

Romaji has several correct answers — and a new official one

Writing Japanese in the Latin alphabet sounds like a lookup, and a chart of a hundred-odd kana makes it look like one. It is not, because more than one system is in real use and they disagree. The same word can be maccha, matcha or mattya, and a person's name can appear one way in a library catalogue and another way in their passport. None of those spellings is a mistake. Each follows a rulebook, and the rulebooks were written for different jobs. This tool prints all of them side by side so you can see exactly where, and why, they part company.

The official standard changed on 22 December 2025

For more than seventy years the Japanese government's own spelling rules were a 1954 cabinet notice built on Kunrei-shiki: し as si, ち as ti, つ as tu, ふ as hu. Almost nobody outside a classroom wrote that way — road signs, passports, company names and textbooks for foreigners had long used Hepburn's shi, chi, tsu, fu. In August 2025 the government's advisory council on the national language recommended bringing the rules into line with that practice, and on 22 December 2025 a new cabinet notice replaced the 1954 one. The official spellings are now Hepburn-based.

For seventy-one years the government's official spelling of ち was ti. Since December 2025 it has been chi — the spelling almost everyone was already using.

Where "Hepburn" disagrees with itself

The new standard is Hepburn-based, not identical to every Hepburn you will meet. Three differences matter. A doubled ち is written cchi in the notice — its own examples are nicchoku and maccha — while dictionaries and library catalogues following the ALA-LC tables, and Japanese passports, write tchi. The syllable ん is always n in the notice (shinbun), where older Hepburn and passports write m before b, m and p (Shimbashi). And long vowels may take a macron (Tōhoku) or be written with the vowels doubled as they are in kana (Touhoku) — the notice permits both. It also says plainly that established spellings such as Tokyo, judo, matcha and Shimbashi, and people's own names, need not change.

Passports follow their own table

The Ministry of Foreign Affairs publishes the table used for passport names, and it is stricter still. Long O and U are not written at all, so 伊藤 is ITO and 大野 is ONO — except that a surname ending in a written お keeps both letters, as in SENOO. The long mark ー disappears (ニーナ is NINA). Loanword sounds are spelled kana by kana, so ティ is TEI and ファ is FUA, and V is not allowed: ヴァ becomes BA or BUA. A spelling such as OHNO is possible only by prior arrangement.

What kana alone cannot tell you

Two decisions depend on the word, not the letters. とうきょう is one long vowel twice, but おもう ("think") ends in two separate vowels, and こうし is kōshi as "lecturer" but koushi as "calf". No rule over kana can separate them, so the tool reads pairs as long by default, lists every pair it merged, and lets you switch. The same goes for は and へ, which are wa and e only when they are grammatical particles — the tool flags them rather than guessing.

Kunrei-shiki and Nihon-shiki are not wrong, just retired

Kunrei-shiki remains the basis of the ISO 3602 standard and fills half a century of Japanese school materials and official documents, so you will keep reading it. Its logic is tidy: the T row is ta ti tu te to because in Japanese those are one series. Nihon-shiki goes a step further and spells ぢ, づ and を as di, du and wo, which makes it the only one of these you can convert back to kana without ambiguity.

For learners across Southeast Asia

Much of the world's Japanese is learned in this region. The Japan Foundation's 2024 survey counted about four million learners outside Japan, with Indonesia second only to China at 732,914 and Thailand and Vietnam also in the top ten. A learner in Jakarta, Bangkok or Ho Chi Minh City meets every system on this page: Hepburn in the textbook, Kunrei in older Japanese-made materials, passport spellings on a colleague's documents, and now the 2025 standard on new signs. Knowing which rulebook a spelling follows is what stops one correct answer being marked wrong against another.

Advertisement

The notice that retired ti, the passport that forbids V, and the letter Hepburn disagrees with itself on

01

On 22 December 2025 a cabinet notice replaced the 1954 romanization rules. The new rules are Hepburn-based: し is shi, ち is chi, つ is tsu, ふ is fu.

02

The 1954 notice it abolished had been issued on 9 December 1954 and was built on Kunrei-shiki, which writes those same four kana si, ti, tu and hu.

03

The 2025 notice writes a doubled ち as cchi — its own examples are nicchoku and maccha — while ALA-LC Hepburn and Japanese passports write tchi.

04

The notice writes 新聞 as shinbun, but lists Shimbashi, tempura and matcha among established spellings that need not change.

05

Long vowels may be written with a macron (Tōhoku) or with the vowels doubled as in kana (Touhoku). The notice permits both, and a circumflex where needed.

06

On a Japanese passport 伊藤 is ITO and 大野 is ONO — long O and U are not written — but 妹尾 keeps both letters: SENOO.

07

The passport table spells ティ as TEI and ファ as FUA, and expressly does not permit VA, VI or VU: ヴァ is BA or BUA.

08

Under the 2025 notice the particles は, へ and を are written wa, e and o — the same letters as the sounds, not the kana.

09

Nihon-shiki keeps ぢ and じ apart as di and zi, and づ and ず as du and zu, which makes it the only system here that converts back to kana unambiguously.

10

The Japan Foundation's 2024 survey counted about four million learners of Japanese outside Japan; Indonesia had the second-largest number, 732,914.

Frequently asked questions

  • Since 22 December 2025 it is the cabinet notice 令和7年内閣告示第4号, which is Hepburn-based: shi, chi, tsu, fu, ji, sha, cha, ja, with ん always written n and long vowels marked with a macron or written with the vowels doubled. It replaced the 1954 Kunrei-based notice. The first row of this tool follows it.

  • No — it is retired as the government standard, not wrong. It is still the basis of ISO 3602 and appears throughout decades of Japanese documents and school materials. If a text is written in Kunrei-shiki, its spellings are correct by its own rules.

  • Both are correct in different systems. The 2025 notice writes a doubled ち as cchi, so maccha; ALA-LC Hepburn and passports write tchi, so matcha. The notice itself lists matcha among established spellings that do not need to change.

  • Because the Ministry of Foreign Affairs passport table does not write long O or U at all. 大野 is ONO and 伊藤 is ITO. A surname that ends in a written お keeps both letters, as SENOO does. A long-vowel spelling such as OHNO is possible only by arrangement with the passport office.

  • Because kana alone cannot show whether おう is one long vowel or two separate ones. In とうきょう it is long; in おもう ("to think") it is not. The tool reads such pairs as long by default and lists each one it merged — switch "Long vowels" to two separate vowels to get omou.

  • は is ha as part of a word and wa only as the topic particle; へ is he in a word and e as a particle. The tool cannot see grammar, so it writes the letter and flags that the input contains は or へ. Edit those by hand if they are particles.

  • No. A kanji's reading depends on the word — 生 alone can be sei, shō, nama, ki, iki and more — so any automatic reading would be a guess presented as a fact. Type the reading in kana and the tool will convert it exactly.

  • The 2025 notice and ALA-LC write ん as n everywhere, so shinbun. Older Hepburn and passports write m before b, m and p, so shimbun and Shimbashi. The notice names Shimbashi as an established spelling that may stay.

  • Hepburn spells for an English reader's ear — shi, chi, tsu, fu. Kunrei-shiki spells by the structure of the kana rows — si, ti, tu, hu — so each row keeps one consonant. The two agree on most syllables and differ on eighteen — し, ち, つ, ふ, じ, ぢ and the twelve combinations built on them — plus how they mark long vowels (macron against circumflex).

  • No. The conversion rules run in your browser; the tool makes no network request with your input and stores nothing.

Method & sources

How it computes

Kana are tokenised (yōon, loanword combinations, っ, ん, ー) and spelled per system: 令和7年内閣告示第4号 本表 and 添え書き 1-4 for the official rows; the modified Hepburn rules followed by ALA-LC; the Ministry of Foreign Affairs passport Hepburn table and its notes; the 1954 notice 第1表 (Kunrei-shiki) and 第2表 spellings (Nihon-shiki) as listed in the 2025 notice's 対照表.

What this tool implements

  • The official rows follow 令和7年内閣告示第4号 (22 December 2025) exactly, and the test suite is built from the notice's own examples: anman, kanpai, shinbun; zasshi, nicchoku, maccha; kāsan/kaasan, Tōhoku/Touhoku; tan'i, en'yūkai; ko'uta.
  • ⚠️ "Hepburn" disagrees with itself and the page shows it. The 2025 notice writes a doubled ち as cchi; modified Hepburn (ALA-LC) and passports write tchi. The notice writes ん as n everywhere; passports write M before B, M and P.
  • The passport row follows the Ministry of Foreign Affairs table and its worked names: long O and U not written (ITO, ONO, YUKI), a final written お kept (SENOO), ー omitted (NINA), loanword kana spelled one by one (TEI, FUA, DEYU), ヴ as BU — V is not permitted.
  • Kunrei-shiki and Nihon-shiki spellings are the 1954 notice's 第1表 and 第2表 as listed in the 2025 notice's 対照表. Neither system defines loanword combinations, so none are invented: those rows spell each kana on its own and the page says why.
  • ⚠️ Kana alone cannot say whether おう is one long vowel (東北 Tōhoku) or two (思う omou). The default reads such pairs as long, lists every pair it merged, and offers the reading with separate vowels — marked with the notice's apostrophe in the official rows. は and へ are written as letters and flagged, because only grammar makes them wa and e.
  • Kanji are refused rather than read: a reading depends on the word.

Sources

What can make this go out of date

  • The loanword combinations are the same set and spellings as kana-chart-kernel.js's EXTENDED group, so the two tools cannot disagree.
Advertisement
Pre-footer · AD-W3 728 × 90