Japanese text you enter is analysed into short units by MeCab and UniDic, and romanised with word division on the basis of the kana form of each unit. The grammatical information that word division requires — telling particles and auxiliaries apart, identifying prefixes and suffixes, recognising proper nouns — is taken from the result of the morphological analysis.

The work is divided as follows.

• Readings follow the result of the analysis. The romaniser does not, as a rule, override them.
• Spelling — the correspondence between kana and roman letters — can be chosen from two standards on the input page.
• Word division always follows ALA-LC §6, whichever spelling you choose.

• ALA-LC Japanese Romanization Table (2022 version) — the romanisation rules of the Library of Congress, widely used in library catalogues. This is the default.
• Rōmaji no tsuzurikata (Cabinet Notification, December 2025) — the Japanese Cabinet Notification of December 2025, broadly based on modified Hepburn.

The two agree for the most part. They differ in the following respects.

Because the Cabinet Notification places sounds used only in loanwords outside its scope (preamble 4), those are converted using ALA-LC chart B even when the Notification is selected. And because it says almost nothing about word division — only that a hyphen may be used (note 5) — word division always follows ALA-LC §6.

• Syllabic n is always written n. 新橋 Shinbashi, 天ぷら tenpura, 欄間 ranma — never m.
• The particles は, へ and を are written wa, e and o.
• シ shi, チ chi, ツ tsu, フ fu, ジ ji. ヲ is o. ヂ and ヅ are ji and zu.
• A long i is written ii. 新潟 Niigata, 兄さん niisan.
• ei is written ei. 平成 Heisei, 時計台 tokeidai, 経済的 keizaiteki, 英語 Eigo.
• An apostrophe follows n before a vowel or y. 翻訳 hon'yaku, 単位 tan'i.
• Proper nouns and the first word of a sentence are capitalised.

• The spelling follows the kana form given by UniDic. The pronunciation form is used only to decide whether a sequence of vowels has merged into a long vowel. This is why 労働 comes out as rōdō while 子牛 comes out as koushi (ko'ushi under the Cabinet Notification). Transcribing the pronunciation form directly would make 子牛 kōshi as well.
• Even though the pronunciation form of 政治 is セージ, it is romanised seiji. Both standards write ei as ei — compare ALA-LC's keizaiteki and Eigo with the Cabinet Notification's Heisei and tokeidai.
• Where the kana form and the pronunciation form diverge by more than the presence of a long vowel, the pronunciation form is taken as the basis, since it represents the modern reading (てふてふ chōchō, あふ坂 Ōsaka). ALA-LC §2.3.2 prescribes modern readings for historical kana usage.
• A geminate is written as a doubled consonant. Where a short unit ends in a geminate (帰っ + て), the consonant of the following unit is supplied when the two are written together: kaette.

The short-unit boundaries given by UniDic are the starting point. Joining, separation and hyphenation are then decided by the rules of ALA-LC §6.

• Particles are written separately (§6.8.4): 幸福への道 kōfuku e no michi.
• Auxiliaries, prefixes and suffixes are joined to the adjacent word (§6.5, §6.6, §6.8.1.1): 支配する shihaisuru, 経済的 keizaiteki, 新幹線 shinkansen.
• An auxiliary verb following the te-form, and auxiliaries of potential or respect, are written separately (§6.8.1.2, §6.8.1.3): 生きていた ikite ita, 我慢出来ない gaman dekinai.
• Jurisdictions, stations and harbours, titles following personal names, and single characters suffixed to proper nouns are hyphenated (§6.9.2.3.2, §6.9.5, §6.10.1, §6.10.6): 東京都 Tōkyō-to, 井上さん Inoue-san, 日本的 Nihon-teki.
• A single-character element before or after a word is joined (§6.2.3, §6.6.1, §6.4.1.1.1): 日本語 Nihongo, 核戦争 kakusensō, 都道府県 todōfuken. But 編, 抄, 考 and 展 are written separately (§6.6.2.1): 君が代考 Kimigayo kō.
• A modifier written in kana or read with kun'yomi is separated (§6.3.1, §6.4.1.1.2): 米騒動 kome sōdō, 水資源 mizu shigen, 春夏秋冬 haru natsu aki fuyu.

Where the rules turn on on'yomi versus kun'yomi (§6.2.3 against §6.3.1, §6.4.1.1.1 against §6.4.1.1.2), the decision is made from the word origin recorded by UniDic: 漢 Sino-Japanese, 和 native, 外 foreign.