Hepburn, and the system you were taught instead
Two romanisation systems are in daily use in Japan. Hepburn writes し as shi, つ as tsu and ふ as fu, following how the sounds land on an English ear. Kunrei, taught in Japanese primary schools and registered as an ISO standard, writes the same three as si, tu and hu, which keeps the kana table tidy at the cost of misleading anyone reading it as English.
Hepburn is what you see outside the classroom: passports, station signs, road signs and almost every dictionary aimed at learners. This tool uses Hepburn throughout. If a textbook shows you tunami and huton, it is following Kunrei, and the words are the ones you already know.
Long vowels, and why passports look different
A long vowel is one syllable held for two beats. Textbooks mark it with a macron — Tōkyō, rāmen — but a macron cannot be typed into most forms and is not accepted in a passport.
The three options here cover what people actually do. The macron is right for study notes and anything typeset properly. Doubling the vowel keeps the length visible where a macron is unavailable. Dropping it is what Japanese passports do by default, which is why 大野 is written ONO and not ŌNO, and why Tokyo has no marks on any airline booking.
Japanese passport holders may apply to write a long O as OH instead — OHNO, OHTA — but that has to be requested, and once chosen it has to match every ticket and visa afterwards.
These options act on the ー mark, which only appears in katakana. A long vowel spelled out in hiragana — the う in とうきょう — is left as it stands, because whether it is one held sound or two syllables depends on the word: 思う is omou, never omō.
The apostrophe after n
ん is written n, but before a vowel or y that leaves the reading ambiguous. Written shinyuu, a reader cannot tell whether the syllables are shi-nyu-u (親友, close friend) or shin-yu-u. Hepburn resolves this with an apostrophe: shin’yuu. This tool adds it automatically wherever the ambiguity arises.
One place it does not follow: Japanese passports write ん as M before b, m and p, which is why 難波 appears as NAMBA. The output here uses N throughout, since that is the modern convention everywhere except the passport office.
Small tsu
っ doubles the consonant that follows it: 学校 becomes gakkou, 切手 becomes kitte. The exception is ch, where the doubling is written as t — 抹茶 is matcha, not cchatcha or macha.
Going the other way, a doubled consonant in romaji comes back as っ, and tch comes back as っ followed by ち, so typing matcha returns まっちゃ.
Why kanji stay as they are
Every kanji has several readings and only the surrounding words decide which one applies. 生 alone can be sei, shou, nama, i, u, ha or ki. Choosing correctly means recognising the word, not the character, and a lookup table cannot do it.
Anything that is not kana passes through untouched, so a line of mixed writing comes back with its kanji in place and the kana around them converted. To romanise a word written in kanji, look up its reading first and type that reading in kana.