How to Type Small Tsu and Other Japanese Typing Exceptions

Most romaji guides stop at the kana chart, and the kana chart is the easy part. か is ka. し is shi. You type enough of those and you start to believe Japanese input is just spelling.

Then you meet きって (kitte, stamp) and type kite, and get きて (kite, to come) — a different word. You meet しんいち (Shin'ichi, a name) and type shinichi, and get しにち (shinichi, a meaningless sequence). You meet コーヒー (koohii, coffee) and type kohi, and get コヒ (kohi, also meaningless).

Nothing is broken. These are the exceptions, and there are far fewer of them than the mess suggests. This page is the reference for all of them.

The two romaji, in one paragraph

There are two romaji, and confusing them is why these rules feel arbitrary. The romaji you type is what you press to get a kana out of an input method. The romaji you read is how the word is pronounced. To produce は you press ha, but when は is a particle it is read "wa".

Everywhere except the particles the two agree, so you can ignore the distinction until は (wa, topic particle), へ (e, direction particle) or を (wo, object particle) show up. The full explanation, including why これは (kore wa, this is) is read kore wa but typed koreha, is in how to type in Japanese.

Everything below is about the typed romaji — the keystrokes. Here is the whole rule set.

Small っ: double the consonant that follows

A small っ is a one-beat pause, and in romaji you represent it by doubling the first letter of the next syllable.

Two-syllable words are where beginners first notice it. The verb 待つ (matsu, to wait) becomes 待って (matte), with a doubled t that exists nowhere in the dictionary form.

The ch case: doubling becomes cch

っ before a ち-syllable does not double the c, it doubles it into cch. This is the single most common place people get stuck, because the sound "tch" suggests a t.

There is no t anywhere in 出張's romaji. Learn it once and every っち word follows.

っ with no consonant after it

A っ at the end of a word, or before a vowel, has nothing to double. Type it directly as xtsu or ltsu.

That second example is genuinely awkward, which is the point: っ before a vowel is rare, and when it appears you spell the small kana out. The x prefix is not a special case for っ — it is the general way to type any small kana, and there is a full table of them further down this page.

ん: the n versus nn trap

Type nn for ん. Always. It is never wrong, and you never have to think about what comes next.

The trap is real, though, and it catches everyone once. A single n followed by a vowel or a y is read as the start of the next syllable, not as ん.

The fix is the same in every case: shinnichi, honnyaku, konnichiha. Some input methods also accept an apostrophe as a separator — shin'ichi and hon'yaku — but nn is universal across every IME, so there is no reason to remember a second option.

Where you can safely use one n is before a consonant other than y: しんぶん (shinbun, newspaper) and かんたん (kantan, easy) work fine with a single n. But しんぶん also works as shinnbun, which is why the practical rule is just to always double it.

Long vowels

Japanese has long vowels that are held for two beats, and the input method reflects the kana, not the sound.

The ー mark: repeat the vowel

So kohi gives you コヒ (kohi) — two short kana, and not a word. You need koohii.

The o-row plus う pattern

A long o in native Japanese words is usually an o-row kana followed by う, and you type it as ou.

You will see Tokyo written Tokyo, Tōkyō and Toukyou in different places. Same sound, three conventions. ou is the one that matches your keyboard, because it matches the kana とう.

Two related patterns worth knowing:

Note that せんせい is typed sensei, not sensee, even though the second e is long. Again: the keystrokes follow the kana, not the pronunciation.

Small kana with x or l

Prefix a small kana with x or l and both work identically. This is how you type the small kana that are not part of a digraph.

In practice you will only need a few of these. ふぁ (fa), ふぃ (fi), ふぇ (fe) and ふぉ (fo) appear in loanwords constantly, and ヴ (vu) is worth knowing for names. In real text, small ゃ ゅ ょ are almost always produced by typing the digraph directly — kya, kyu, kyo — so you never touch the x prefix for them.

Alternative spellings: all of these work

Japanese has competing romanization systems, which is annoying for textbooks and convenient for typing, because input methods accept the spellings from several of them. Every pair below produces the identical kana.

Pick one column and stay in it. Mixing them is harmless for the machine and confusing for your own muscle memory.

The three rare kana: づ, ぢ and を

These three do not follow from anything you already know, so learn them as a set.

づ and ぢ are rare because modern Japanese pronounces them the same as ず (zu) and じ (ji) — 三日月 (mikazuki, crescent moon) and 縮む (chijimu, to shrink) sound like they contain ず and じ. But the kana differ, so the keystrokes differ. zu gives you ず; du gives you づ.

を is the one particle that does not diverge in pronunciation. It is typed wo and read "wo" — the only oddity being that it is almost always a particle and almost never a word.

Three impossible-looking words, taken apart

Rule-by-rule decomposition is the whole skill. Here is the canonical one, 出張 (shucchou, business trip):

Result: shucchou. Now the same treatment for ちょっと (chotto, a little):

Result: chotto. And one with three rules at once — コーヒー (koohii, coffee): コ is ko, ー repeats the preceding vowel so オ becomes o, ヒ is hi, and the second ー repeats the i, giving i. Result: koohii.

Every word that looks impossible falls apart the same way. Break it into kana, apply the rules, reassemble. If a word still does not come out right, the culprit is almost always ん or っ.

How these rules work on Romaji.live

Here is the twist that matters for practice. On Romaji.live you type the pronunciation romaji, not the IME keystrokes — because the goal is to internalize the sound of the sentence, not to operate a Japanese input method.

That means the particles behave differently here than they do in an IME:

Small っ and long vowels come out the same either way — きって is kitte, コーヒー is koohii — because doubling and vowel repetition are features of the pronunciation, not of the input method. The divergence is confined to the particles.

If you have not yet met the particle は as a grammar point, particles covers what they do; all you need here is the spelling.

The one-page reference

Bookmark this page. You will come back to it for the first month and then never again.

The fastest way to make these rules automatic is to type a hundred sentences that contain them, and Romaji.live is built for exactly that — real sentences, one word at a time, with the romaji shown until you stop needing it. Start with the kana table if the base syllables are still shaky, then move into practice. If a rule looks correct here but your input method refuses to cooperate, the problem is probably the setup rather than the rule: see Japanese IME troubleshooting. And when you hit a word whose reading you cannot work out by hand, paste it into the romaji converter — it will show you the standard romanization, which is usually enough to reveal which rule applies.