Translate a JSON i18n file into Japanese

What changes when a UI goes into Japanese

Japanese mixes three scripts (kanji, hiragana, katakana) and puts no spaces between words. A Japanese string often has fewer characters than its source, but each one is full-width, and loanwords in katakana grow quickly: “Download” becomes ダウンロード, “Account settings” アカウント設定. A button sized for English can overflow or look half empty.

Punctuation changes too: the comma 、, the full stop 。, the full-width question mark ？. “Saved!” typically becomes 保存しました。.

• Line breaking: with no spaces, browsers break between characters. Set lang="ja" and try line-break: strict so that 。 or 、 never starts a line.
• Fonts: kanji share code points with Chinese. Without a Japanese font in your font-family stack and lang="ja", some characters render in their Chinese form.

These are CSS settings: Manuce only works on the string file.

Plurals, placeholders and word order

Japanese has a single CLDR plural category, other, so there are no categories to add, unlike Polish or Arabic. In i18next, only items_other is used for Japanese; an items_one key inherited from the source stays in the file, since keys are never changed, but Japanese plural rules never pick it. In ICU, {count, plural, other {#件のメッセージ}} is enough; only exact matches such as =0 or =1 still apply.

The real trap is word order: the verb goes last, so the placeholder moves and takes a particle.

• "Delete {{count}} files?" becomes "{{count}}件のファイルを削除しますか？", with a counter word (件, 個, 人) that depends on what is being counted.
• "Welcome back, {name}" becomes "おかえりなさい、{name}さん": the honorific さん follows the variable.

A sentence split across keys and joined in code (t('sentBy') + name) cannot be reordered for Japanese. Merge it into one value with a placeholder before uploading.

Register: buttons, messages and keigo

A professional Japanese interface uses the polite -masu / -desu form for sentences and short nouns for actions: 保存 or 送信 on a button, パスワードが正しくありません。 for an error rather than the blunt plain form 正しくない.

Switching register from one screen to the next stands out at once. Check it on the free sample: one paragraph of your file translated at the level you choose, before you pay. For an interface aimed at Japanese customers, Professional (domain specialist) or Premium (most advanced engine, then cross-review) fits best; Essential is for understanding or internal testing.

For brand and product names, check on the sample and then on the delivered file whether they stay in Latin script or move to katakana, in line with your market's usage.

What Manuce returns, and what it does not do

You get the same file back: indentation, key order, quotes, comments, line endings, encoding and BOM are preserved. Only the values are now in Japanese. Flat or nested JSON and arrays of strings are supported.

• Every value is compared with the source: {name}, {{count}}, %s and ICU messages (plural, select) must appear identically, the same number of times. If a mismatch persists, the source value is kept and the report says so.
• Not translated: keys, empty values, URLs, email addresses, numbers, colours, date formats, identifiers, and anything marked as non-translatable.
• The report lists translated strings and those left unchanged, with the reason.

Manuce does not edit your CSS or components, convert the encoding or adapt date patterns: "MM/DD/YYYY" stays as it is, whereas Japan writes 2025/03/14 or 2025年3月14日. Use Intl.DateTimeFormat('ja-JP') or change that value yourself. Pricing covers the words in the values, keys and tags excluded, with a ×1.6 coefficient for Japanese.

Preparing the file and checking the result

1. Encoding: the original encoding is kept. A Latin-1 file cannot hold Japanese; save it as UTF-8 before uploading.
2. Concatenation: merge sentence fragments into full values with placeholders, so Japanese word order is possible.
3. Counter context: a bare count value says nothing about what is counted; a full value such as "{{count}} files" allows the right counter word.
4. After delivery: read the report, then run the app in ja and check buttons, narrow menus and line breaks.

In practice

1. Save the JSON file as UTF-8 and merge sentences split across several keys.
2. Upload it: the quote appears straight away, based on the words in the values.
3. Check the free sample: polite register, 、。 punctuation, placeholder position.
4. Download the Japanese file and its report, then put it back in your locales folder.
5. Test the UI in ja: overflow, line breaks and Japanese font rendering.