Romaji to Hiragana: How Japanese Input Conversion Works

You press s, a, k, u, r, a and さくら appears. Nothing in that sequence is mysterious once you know there is a small program in the middle doing the work — and knowing what that program does is what lets you predict what your keyboard will produce.

This article is about the conversion layer itself. For installing the keyboard on your machine, see Japanese keyboard setup.

What an IME actually does

Typing Japanese means typing through a converter rather than typing characters directly.

An IME — Input Method Editor — sits between your keystrokes and your document. It watches the letters you press, matches them against a romaji-to-kana table, and decides what to emit. The pipeline has three stages:

1. You type letters. k, a
2. The IME converts them to kana. か
3. You optionally convert kana to kanji. か → 加 / 家 / 可 / 歌 …

Stage 2 is automatic and instant. Stage 3 is manual, and it is the part that surprises people, because Japanese has far too many homophones for the IME to guess. こう could be 高, 校, 光, 効, 講, 向, 幸, 港, 功, 交, and a few dozen more. The IME cannot know which you mean, so it shows you a list and waits.

Stage 2: letters to kana

This is the conversion that runs on its own as you type. It follows the romaji values of the kana chart:

Your IME is doing something slightly cleverer than a lookup table, though. It processes letters greedily and looks ahead, which is why n behaves the way it does.

Why ん is the awkward one

Typing n produces ん immediately — unless the next letter would form a valid syllable starting with n. So:

• kane → かね (the n is the start of ね)
• kann → かん (the second n confirms the first was ん)
• kan alone → かん (nothing follows, so ん is assumed)

That lookahead is why nn is the safe spelling for ん and single n is only safe before a consonant. しんいち typed as shinichi gives しにち, because the IME reads ni as a valid syllable. You need shinnichi or shin'ichi. Full details in romaji typing exceptions.

Alternative spellings all work

Because Japanese has competing romanization systems, IMEs accept several spellings per syllable. All of these work in every mainstream IME:

Pick one style and stay with it until it is automatic. Mixing si and shi randomly is how typos happen.

Small kana

The double-consonant form is the one that matters in practice — it is how っ appears in real words, and it is covered in detail in romaji typing exceptions.

Stage 3: kana to kanji

Once you have kana on screen, press Space to convert. The IME offers candidates; press Space again to move through them, or use the number keys to jump straight to one. Enter confirms.

Two habits make this much less annoying:

• Type the whole phrase before converting, not word by word. IMEs are far more accurate on longer chunks because they can use the surrounding grammar. こうこうにいきました will convert correctly as a unit; converting each piece separately is what produces nonsense.
• Enter confirms, and confirming is what strands a wrong choice. If you confirm 校 when you meant 高, you cannot fix it by pressing Space again — the text is now committed. Backspace and retype, or use the IME's reconversion (varies by IME; often it is select-then-convert).

When you genuinely only want kana

Press Enter instead of Space to keep what you typed as hiragana. For katakana, there is usually a dedicated key or shortcut — on Windows IME it is typically the Kana key on a JIS keyboard or Alt + Caps Lock, on macOS the Kana input source or the katakana conversion in the candidate list. For a word you actually want in katakana, just convert and pick the katakana candidate.

The mode indicators, and why they matter

Every IME has modes, and almost every "my keyboard is broken" report is really a mode problem.

That あ / A toggle is the one you will use constantly — it is how you write an English word inside a Japanese sentence without switching input sources entirely. On Windows the toggle is typically the 半角/全角 key (or Alt + ~); on macOS there is a shortcut you can set in Keyboard settings.

There is a second trap hiding here: full-width vs half-width Latin. If you type in 全角 (full-width) mode, this becomes ｔｈｉｓ — Latin letters that look normal-ish but are not ASCII, and which will break search boxes, URLs and code. If your text looks subtly wrong, check that the half-width (半角) mode is on. This is covered with fixes in IME troubleshooting.

Will this interfere with my normal typing?

No. The IME is an input source, not a global setting — it only acts when you switch to it. Your English typing is untouched, and you can have Japanese and English active side by side and toggle between them mid-sentence.

The one thing worth setting up deliberately is the toggle, because you will use it constantly. Find a shortcut that is easy to reach and learn it now; a comfortable switch is the difference between typing Japanese feeling fluid and feeling like a chore.

Practice the conversion, not just the characters

Converting is a skill with a specific shape: see the kana, retrieve its romaji, produce it with your fingers, without pausing to think. That retrieval is what makes it automatic, and it only comes from repetition on real words and sentences rather than on isolated characters.

That is the whole idea behind romaji typing practice — sentences in, romaji out, immediate feedback.

On Romaji.live you type the pronunciation romaji, not the IME keystrokes — so これは is practiced as kore wa even though your IME wants koreha. Practicing the spoken form trains your reading; the keystroke form is what keyboard setup gets working. Different jobs, and it is worth being deliberate about which one you are drilling. If you want to see how a given sentence turns out under the standard reading — particles included — the romaji converter will spell out any Japanese you paste in.