How to Type Japanese (IME): Setup, Romaji Input, and Shortcuts
Typing Japanese on any modern operating system comes down to three moves: install an Input Method Editor (IME), switch the keyboard into Japanese mode, and type romaji that the IME converts into kana and kanji.12 The romaji you type is a typing-only convention called wāpuro rōmaji. It is standardized at the input layer by JIS X 4063:2000, and it is not the same as the romanization systems you read.34
The Short Answer
Install the IME, switch the input mode, type romaji
The setup follows the same three steps on every supported platform.
- Add Japanese as a system input source in the operating system's keyboard or language settings.125
- Switch the active input layout to the Japanese IME with the OS-level shortcut: Windows logo key + Spacebar on Windows,1 Control + Spacebar (or Control + Option + Spacebar for the next source) on macOS,6 and the globe key on iOS.5
- Type Latin letters. The IME shows kana under a composition underline, and Spacebar opens the conversion candidate window.1
On Windows, the system tray shows a circled-J badge plus a mode indicator: あ when composing kana, A when in half-width alphanumeric.1 On macOS, the menu-bar Input menu shows two installable Japanese sources, "Japanese - Romaji" and "Japanese - Kana". Learners on non-JIS hardware want Romaji.2
The tray badge sequence is your status line: Ⓙ means the Japanese IME is the selected layout, あ means the next keystrokes will compose hiragana, and A means alphanumeric pass-through.1 If あ is missing, the IME is installed but still in alphanumeric mode.
A first commit, in two lines:
寿司7
"sushi"
Typed input trace: s u s h i produces すし in the composing buffer. Spacebar shows 寿司 in the candidate window. Enter commits.18
Why this is called wāpuro rōmaji and why it matters
The romaji you type into an IME is wāpuro rōmaji (ワープロローマ字), short for wādo purosessā rōmaji, "word-processor romaji." Its more formal Japanese name is rōmaji kana henkan (ローマ字仮名変換), "Roman-character kana conversion."4
The convention is standardized by JIS X 4063:2000, Keystroke to KANA Transfer Method Using Latin Letter Key for Japanese Input Method, published by the Japanese Standards Association.3 The standard text is explicit that it specifies a keystroke-to-kana conversion behavior, not a romanization of Japanese.34
In practice, wāpuro accepts both Hepburn-style keystrokes (shi, tsu, chi, ji) and Kunrei-shiki or Nihon-shiki keystrokes (si, tu, ti, zi) for the same kana. It also adds IME-only spellings (the l- and x- small-kana prefix family, the nn terminator for ん) that no published romanization system uses.849 The reading-side framework is covered in Romaji Explained: Hepburn, Kunrei-shiki, and Nihon-shiki.
Wāpuro is a one-to-one transcription of the kana, not of the spoken syllables. The particles は, へ, and を are typed ha, he, and wo, matching the spelling, not the pronunciation /wa/, /e/, /o/.4
How an IME Works (the Conceptual Flow)
Text in, kana out, kanji on demand
An IME inserts a four-stage pipeline between the keyboard and the focused application.1 The default flow uses explicit conversion, and every platform shares it.
These stage names are worth knowing.
- 変換 (henkan) is conversion, the candidate-selection step.10
- 無変換 (muhenkan) means "no conversion." It is a dedicated JIS-keyboard key that commits the composing buffer as kana without opening the candidate window.10
- 確定 (kakutei) is commit, the final-acceptance step. On Microsoft IME, it maps to Enter and ends the conversion process.110
Microsoft Learn documents both ローマ字入力 (romaji input) and かな入力 (kana input) as modes users can select on the IME toolbar.1 Romaji input is the default for users on non-JIS hardware.2
macOS has a Live Conversion setting that converts hiragana into kanji as you type, without an explicit Spacebar press.11 The four-stage pipeline above describes the default explicit-conversion path that all IMEs share. With Live Conversion on, stages two and three collapse and the visible behavior changes.
The two states: composing vs committed
A composing buffer is text the IME still owns; a committed string is text the application owns. The Windows IME marks the composing state with an underline.1
Until you press Enter or convert the next segment, the F-key conversion shortcuts (F6 through F10) apply to the current composing buffer.1 After 確定, the text is delivered to the application and the F-keys return to their application-level meanings.1
The Microsoft Learn IME reference table assigns the F-keys these meanings:1
- F6: Convert input string to hiragana.
- F7: Convert input string to full-width katakana.
- F8: Convert input string to half-width katakana.
- F9: Convert input string to full-width alphanumeric.
- F10: Convert input string to half-width alphanumeric.
Set Up Japanese Input on Your OS
Windows (Microsoft IME)
The conceptual path is Settings → Time & language → Language & region → Add a language → Japanese → Install.1 Menu labels change across Windows versions, so the destination matters more than the exact click sequence.
After installation, the Japanese IME includes both hiragana mode and half-width alphanumeric mode. The system-tray indicator toggles between あ (hiragana) and A (alphanumeric).1
The OS-level switch is Windows logo key + Spacebar, which cycles between installed input methods.1 Inside the IME, Alt + ` (back quote) flips between Japanese composing mode and direct English input without leaving the IME.1 Kana-input mode, the alternative for JIS-hardware typists, toggles with Alt + カタカナ/ひらがな/ローマ字 when the corresponding right-click setting is enabled.12
macOS
The conceptual path is Apple menu → System Settings → Keyboard → Text Input → Edit → + → Japanese → Japanese - Romaji.2 JIS-keyboard users pick "Japanese - Kana" instead.
Apple documents only two Japanese input sources: Japanese - Kana and Japanese - Romaji.2 Apple describes Romaji as the right choice "when you're not using a Japanese keyboard; text you type is automatically converted to hiragana."2
OS-level cycling on macOS uses Control + Spacebar for the previous input source and Control + Option + Spacebar for the next; holding Control + Spacebar opens the full source picker.6 The Caps Lock key can also be configured to toggle between a non-Latin input source (Japanese) and a Latin input source (English).6
macOS also exposes a Live Conversion setting under Japanese input settings; with it enabled, hiragana converts into kanji as you type, with no explicit Spacebar press.11
iOS
The conceptual path is Settings → General → Keyboard → Keyboards → Add New Keyboard → Japanese.5
Apple's iOS Japanese input offers two layouts.5 The Romaji layout is QWERTY-style: type Latin letters, see kana. The Kana layout is the 12-key flick grid that is the dominant input method among native users on mobile.
To switch between installed keyboards on iOS, tap the globe key (🌐) on the on-screen keyboard, or long-press the globe key to pick from a menu.5
Android (Gboard)
Gboard is Google's keyboard for Android and ships as the default on Pixel devices and many other Android phones. The conceptual approach: open Gboard's language settings, add Japanese, and choose either the QWERTY layout (for romaji input) or the 12-key layout (for flick input).
No single Google support page documents the exact menu sequence for adding Japanese to Gboard in the way Apple documents iOS. On-device menu paths also shift across Android versions and OEM skins. The conversion engine under Gboard's Japanese mode is Mozc, the open-source IME documented in the google/mozc repository.13
Google Japanese Input and Mozc (desktop alternative)
Google Japanese Input (Google 日本語入力) is Google's proprietary freeware desktop IME, first released on 2009-12-03.14
Mozc is the open-source release that "originates from Google Japanese Input," published under the BSD 3-Clause license at github.com/google/mozc.1314 The Mozc README lists Android, Apple macOS, Chromium OS, GNU/Linux, and Microsoft Windows as target platforms.13
The main functional difference from Microsoft IME or Apple's IME is dictionary coverage. Google Japanese Input's dictionaries "are generated automatically from the Internet," extending coverage to personal names, internet slang, brand names, and neologisms.14 Open-source Mozc does not use Google's closed-source algorithms for generating dictionary data from online sources. As a result, it ships a smaller dictionary than the proprietary product even though the conversion engine is the same.14
The Mozc README is also explicit about the tradeoff: "Mozc is not an officially supported Google product," Google "provides no official quality assurance (QA) for the open-source version," and there is no designated stable release.13
The Romaji-Mode Typing Flow
A worked example: sushi to すし to 寿司
The standard compose-then-convert loop uses the same mechanics on every platform.18
- Type
s u s h i. The IME composes すし under a composition underline.1 - Press Spacebar. The conversion candidate window opens; 寿司 appears in the candidate list alongside すし and スシ.1
- Press Enter to commit the highlighted candidate. The underline disappears, and the text is delivered to the application.1
The Microsoft Learn IME documentation uses a similar h a r u ⇒ はる ⇒ 春 trace as its canonical worked example.1
The conversion candidate window
The candidate window opens on Spacebar. Use arrow keys to move, number keys to jump to candidates 1 through 9, Tab to expand the list into a multi-column table, and Enter to accept.1
On macOS, Spacebar opens the same window. Down and Up arrows move between candidates, Return commits, and Escape reverts to the original composing text.16 macOS also has an "alternate candidates" shortcut at Control + Shift + R (or Option + Shift + R on Windows-style keyboards). It shows typographic alternates such as half-width vs full-width digit shapes.16
The Microsoft IME candidate window supports two extra actions worth knowing. Ctrl + Delete removes the highlighted candidate from the prediction history. Ctrl + B searches the highlighted candidate, which is useful for disambiguating kanji compounds.1
How wāpuro romaji differs from reading romaji
Wāpuro is multi-mapping by design. Both Hepburn keystrokes (shi, tsu, chi, ji, fu) and Kunrei-shiki or Nihon-shiki keystrokes (si, tu, ti, zi, hu) resolve to the same kana.84
Wāpuro also adds spellings that no reading-romaji system uses: the l- and x- prefix family for small kana, doubled consonants for っ, nn for final ん, and the hyphen-minus key for the chōonpu ー.849
Long vowels follow the underlying kana. Wāpuro spells long ō as o u for おう and o o for おお, never with a macron, because macron-bearing characters are not on any standard keyboard.4
東京17
"Tokyo"
Typing input: t o u k y o u. The wāpuro convention follows the kana spelling, so each long ō here is two keystrokes.84
Typing Special Characters
Small tsu (っ): double the next consonant, or type ltu / xtu
There are two documented ways to type the small tsu.8
The usual route is doubled consonant: type the next consonant twice. k i t t e produces きって. Apple's wording: "To enter blocked sounds (double consonants), double the first consonant of the next syllable."8
The fallback route is the standalone small tsu, useful when the doubled-consonant pattern does not apply (stylized text, certain transliterations). Type l t u, x t u, l t s u, or x t s u. Apple's roman-to-kana table lists all four spellings.8
切手8
"postage stamp"
Typing input: k i t t e (doubled-consonant route to っ).8
Small ya, yu, yo (ゃ, ゅ, ょ): just type the contracted sound
The direct route is to type the contracted sound itself. kya, shu, and cho produce きゃ, しゅ, and ちょ in one move. The IME composes the small kana automatically.89
The fallback route prefixes with l or x for the rare standalone case. Apple documents xya, lya for ゃ, xyu, lyu for ゅ, and xyo, lyo for ょ.8
学校8
"school"
Typing input: g a k k o u. Doubled-consonant for っ, kana spelling for the long ō.84
The n problem: nn for word-final ん
A lone n is ambiguous, because it could be the start of na, ni, nu, ne, or no. The IME holds it in the composing buffer until the next keystroke clarifies it, which means word-final ん needs an explicit terminator.89
Apple's mapping table documents three valid commit forms for ん: nn, n', and xn.8
The practical rule has three cases.
- Typing
nihonand pressing Spacebar can still resolve to にほん, because Spacebar acts as a commit boundary. - Typing
nihonfollowed immediately by another consonant, as innihonjin, commits the ん automatically when the next consonant arrives. - Typing
nihonn, ornihon'with an apostrophe, always commits the ん in place.8
n without a terminator is the most common new-typist bugThe chōonpu (ー) and punctuation
Type the long-vowel mark ー (chōonpu) with the hyphen-minus key (-) while the IME is in any kana mode.8 The mark is the standard katakana long-vowel device.
Japanese full stop 。 and comma 、 map to the . and , keys by default when the IME is composing in a Japanese mode. The half-width Latin . and , reappear automatically when the IME is in alphanumeric mode.
The precise punctuation map (? to ？, ! to ！, brackets, the middle dot ・) is configurable in Microsoft IME's "Input settings such as punctuation" pane.1 Apple's macOS guide exposes equivalent configuration through Japanese input source settings.11 Apple's published roman-to-kana table focuses on small kana and ん and does not list chōonpu or punctuation rows. The mappings above are therefore taken from the default IME behavior on both platforms rather than from a single Apple reference row.
Small character and punctuation reference table
| Target | Wāpuro input | Notes |
|---|---|---|
| っ (sokuon) | ltu, xtu, ltsu, xtsu, or doubled consonant | Doubled-consonant route is dominant; e.g. kitte ⇒ きって.8 |
| ゃ | xya, lya, or contracted form (kya ⇒ きゃ) | Direct contracted-sound input is the default route.8 |
| ゅ | xyu, lyu, or contracted form | Same as ゃ.8 |
| ょ | xyo, lyo, or contracted form | Same as ゃ.8 |
| ぁ ぃ ぅ ぇ ぉ | xa / la, xi / li, xu / lu, xe / le, xo / lo | Standalone small vowels.8 |
| ゎ | xwa, lwa | Standalone small wa.8 |
| ん (word-final) | nn, n', or xn | Single n holds in the buffer until disambiguated.8 |
| ー (chōonpu) | - (hyphen-minus key) | Active in any kana mode.8 |
| 。 (kuten) | . (period) | Default mapping in Japanese composing mode; configurable on both platforms.111 |
| 、 (tōten) | , (comma) | Default mapping in Japanese composing mode; configurable on both platforms.111 |
Switching, Converting, and Toggling: The Keyboard Shortcuts
OS-level input-source switching
The top-level toggle gets you between English and Japanese without touching the IME's own settings. Windows uses Windows logo key + Spacebar to cycle through installed layouts.1 macOS uses Control + Spacebar for the previous source and Control + Option + Spacebar for the next. Holding Control + Spacebar opens a picker.6 On macOS, Caps Lock can be configured to toggle between a non-Latin source (Japanese) and a Latin source (English).6
On iOS, the globe key (🌐) on the on-screen keyboard cycles installed keyboards. Long-press it to pick from a menu.5 On Android, Gboard follows the same idea with a language key on the space bar or a globe key.
Within-IME mode switching
The inner toggle flips between hiragana mode, katakana mode, half-width katakana, full-width romaji, and half-width romaji without leaving the IME. On Windows, the main route is the F-key conversion shortcuts applied to the current composing buffer.1 On macOS, the F-keys behave the same way. A parallel set of Control + letter shortcuts (Control + J / K / L / Semicolon / Colon) is also available.16
Microsoft IME also offers a within-IME English toggle at Alt + ` (back quote), which flips between Japanese composing mode and direct English input without changing the OS-level input source.1
Converting on the fly: space, F6, F7, F8, F9, F10
The F-keys apply to the current composing buffer, not to committed text.1 The most useful learner case is F7: type a loanword in hiragana by reflex, press F7 instead of Spacebar, and the buffer changes to full-width katakana in one move.
東京17
"Tokyo"
Worked F7 trace: type t o u k y o u. The composing buffer shows とうきょう. Press F7 (instead of Spacebar). The buffer changes to トウキョウ. Enter commits.1817
The Apple macOS guide assigns the same F-key meanings (F6 hiragana, F7 katakana, F8 half-width katakana, F9 full-width romaji, F10 romaji) and pairs them with Control + letter alternatives.16 After Enter (commit / 確定), the F-keys return to their application-level meanings.1
Half-width vs full-width
Unicode Standard Annex #11, "East Asian Width," defines Halfwidth (H) and Fullwidth (F) as two compatibility categories. They pair characters with explicit half-width or full-width compatibility decompositions to characters elsewhere in Unicode.18 The distinction comes from "DBCS mixed-width encodings" (double-byte character sets such as the JIS family). In those encodings, Latin and katakana characters existed in two display widths, with halfwidth occupying a single-byte slot and fullwidth occupying two bytes.18
The Microsoft IME exposes Half-width Alphanumeric (A, 1) and Full-width Alphanumeric (Ａ, １) as distinct input modes. F9 (full-width) and F10 (half-width) switch the current composing buffer between them.1 macOS exposes the same two states as separate input sources and as F9 / F10 conversion shortcuts.16
A form field validating a phone number, postal code, or numeric ID may strip full-width spaces or refuse Ａ, １ while accepting A, 1. Before submitting, commit numeric-looking data in half-width mode (F10 on Microsoft IME1; Control + Apostrophe on macOS, ANSI keyboard16). The category distinction is structural in Unicode, not cosmetic.18
Shortcut reference table (Windows IME, macOS, Google IME)
The right-hand column below records Google Japanese Input and Mozc on Windows. Mozc on Windows targets the same Windows Text Services Framework surface as Microsoft IME. It inherits Microsoft IME's F-key bindings as the de facto Windows-IME convention.113 The bindings are not separately documented in the Mozc repository, so treat the right-hand column as inherited-by-default behavior rather than a source-confirmed map.
| Action | Microsoft IME (Windows) | Apple Japanese IME (macOS) | Google Japanese Input / Mozc (Windows) |
|---|---|---|---|
| OS-level: cycle input sources | Win + Space1 | Control + Space (prev) / Control + Option + Space (next)6 | Win + Space (OS-level, unchanged by IME)1 |
| Within-IME: toggle Japanese / English | Alt + ` (back quote)1 | Caps Lock (when set to toggle Latin / non-Latin)6 | Alt + ` (Mozc inherits Windows-IME convention)113 |
| Hiragana (conversion on composing buffer) | F61 | F6 or Control + J16 | F6113 |
| Full-width katakana | F71 | F7 or Control + K16 | F7113 |
| Half-width katakana | F81 | F8 (or Control + Semicolon in some modes)16 | F8113 |
| Full-width alphanumeric | F91 | F9 or Control + L16 | F9113 |
| Half-width alphanumeric | F101 | F10 or Control + Colon16 | F10113 |
| Open candidate window | Space1 | Space16 | Space113 |
| Commit / 確定 | Enter1 | Return16 | Enter113 |
| Revert / cancel composition | Esc1 | Esc16 | Esc113 |
| Custom-dictionary registration | Right-click IME → Add Word, or User Dictionary Tool12 | macOS user dictionary editor, via Japanese input settings11 | Mozc dictionary tool13 |
Advanced Features Worth Knowing
The custom dictionary: teach the IME your name
The custom dictionary is a word list you can extend. The IME's converter consults it alongside its built-in dictionary. Proper nouns (personal names, place names, company names) often convert badly out of the box because they are not in the default dictionary. The custom dictionary is the fix.
All three desktop IMEs use the same basic schema: reading (yomi) → written form (kanji or kana) → part of speech (品詞, hinshi). The part-of-speech tag lets the converter inflect or place a custom entry correctly.1213
The registration path differs by platform.
- Microsoft IME (Windows): register a word via the system-tray IME icon, right-click → Add Word, or via the User Dictionary Tool surfaced from the right-click menu.12
- Apple Japanese IME (macOS): register a word in the macOS user dictionary, accessible from Japanese input source settings.11
- Mozc and Google Japanese Input: ship a dedicated dictionary tool for adding entries (reading, written form, part of speech).13
Why power users switch to Google IME or Mozc
Dictionary freshness and coverage are the main reasons. Google Japanese Input's dictionaries are "generated automatically from the Internet," giving stronger coverage of personal names, internet slang, brand names, and neologisms than statically curated dictionaries.14
Platform breadth is the second reason. Mozc, the open-source release of the same conversion engine, runs on Android, macOS, ChromeOS, GNU/Linux, and Windows.13 That cross-platform footprint is wider than either Microsoft IME (Windows-only) or Apple's IME (macOS and iOS).
There are two tradeoffs to weigh. Mozc does not ship Google's closed-source web-dictionary generation, so the open-source build has a smaller dictionary than proprietary Google Japanese Input even though the conversion engine is the same.14 The Mozc README also states that Google "provides no official QA" for the open-source release and does not designate stable releases.13
Stroke-input and handwriting fallbacks
When you can recognize a kanji but cannot type its reading, every major desktop IME provides a recognition or browsing fallback.
On Windows, Microsoft Learn documents the IME Pad, which you access from the IME right-click menu. It has five modes: Hand Writing (draw a kanji, select from recognized candidates), Character List (browse a font for the character), Soft Keyboard (on-screen kana / alphanumeric keyboard), Strokes (browse by stroke count), and Radical (browse by radical, then by stroke count).1
On macOS, Apple ships handwriting input as a system-level Trackpad Handwriting source (separate from the Japanese IME) and a Character Viewer. The Japanese IME's candidate window also shows homophone candidates with annotation hints that help disambiguate kanji the user only partially remembers.1916
On iOS, a "Japanese - Handwriting" keyboard can be added from Settings → General → Keyboard → Keyboards → Add New Keyboard.5
Good to know
Typing romaji into an IME is not "using romaji"
Wāpuro rōmaji is a typing convention defined as a keystroke-to-kana conversion method by JIS X 4063:2000. The standard text is explicit that it does not define a romanization of Japanese.34 The reading surface (what the learner sees on screen) is kana and kanji. The romaji exists only between the learner's keystrokes and the IME, and disappears at commit. The "stop reading romaji" rule covered in Stop Using Romaji: When to Switch to Kana Permanently is about the reading surface, not the typing surface.
Long vowel typing diverges from reading romaji
Reading-romaji conventions like Hepburn write long ō with a macron (tōkyō). Macron-bearing characters are not on any standard keyboard and produce nothing useful when typed into an IME. Wāpuro spells long ō by typing the underlying kana sequence: o u for おう, o o for おお.4 The same word that reads tōkyō in a textbook is typed t o u k y o u:
東京17
"Tokyo"
Why your Japanese sometimes looks wrong on a website
Japanese web forms often validate against the half-width Unicode category and silently reject or strip full-width characters.18 A phone number committed as ０９０-１２３４-５６７８ may submit as empty. A postal code in full-width digits may fail validation with no visible error. The fix is to switch to half-width alphanumeric (F10 on Microsoft IME1; Control + Apostrophe on macOS, ANSI keyboard16) before committing anything that looks like a code, ID, or number.
The IME pad and handwriting input are unknown-kanji rescues
When you can see a kanji but cannot type its reading, the Windows IME Pad (Hand Writing, Strokes, or Radical mode) and the macOS Trackpad Handwriting source both let you find the character by shape instead of by sound.119 These fallbacks are slower than typing a reading. But when the alternative is opening a separate reference, they are faster than a radical-based dictionary lookup.
Mobile flick input is what natives use
iOS and Android both ship a 12-key flick layout. Each consonant column is a single key, and up, down, left, and right flicks reach the four vowel slots in that column.5 The QWERTY-style romaji layout is available on both platforms, but it is the default for many learners and non-native users rather than for native speakers. A one-month investment in flick is worth the cost for any learner who reads Japanese on mobile regularly.
The name "wāpuro" is short for "word processor"
When Latin-keyboard kana input first emerged, it lived on dedicated Japanese word-processor hardware (ワードプロセッサ, wādo purosessā), abbreviated ワープロ (wāpuro). The keystroke convention picked up the hardware's name and kept it after the hardware disappeared.4 The more formal name ローマ字仮名変換 (rōmaji kana henkan, "romaji kana conversion") survives in standards documents like JIS X 4063.34
See also
- Hiragana vs. Katakana: How to Tell Them Apart and Use Both
- The Complete Hiragana Chart (Gojūon): How to Read All 46 Base Kana, Dakuten, and Yōon
- The Complete Katakana Chart (Gojūon): How to Read All 46 Base Kana, Dakuten, and Yōon
- How to Write Your Name in Katakana: Foreign-Name Transcription Rules with Examples
- Hiragana, Katakana, or Kanji First? A Beginner's Script Order