👤 Japanese Full Name Splitter

Paste one full name per line. The tool tries whitespace split, then a 300+ surname dictionary, then a 2-character fallback to extract family and given names with a confidence rating. Edit any wrong split inline and export as CSV / TSV. Readings are filled instantly from a built-in 50 KB dictionary.

100% Free No signup Browser-only Opt-in only 5 languages Dark mode

Browser-only by default. The "high-accuracy mode" checkbox enables server-side MeCab processing (no logs).

⚠️ About Accuracy

This tool is not 100% accurate. Compound surnames, rare names, non-Japanese names, and katakana-only names may be misclassified. Always verify manually before final use.

Readings come from a built-in 50 KB dictionary. It runs instantly, but rare proper nouns and uncommon kanji may be left blank.

🔒 About Privacy

・All processing happens entirely in your browser
・Name data is never sent to any server
・No storage logs, no upload history, no database

Full names (one per line)

🎯 High-accuracy mode (MeCab, server-side)

Uses IPA dictionary morphological analysis for more accurate name and reading detection.

⚠️ Checking this sends names to the server (no logs, no database).

Result (editable)

📖 Where people get stuck

Splits a Japanese name into family name and given name and estimates the reading. By default everything runs in the browser against a built-in dictionary; only the high-accuracy mode sends the text to a MeCab process on the server, which keeps no logs. This can never be perfect in principle — neither where to split nor how to read it is determined by the written form alone. The same four characters split differently for different people, and the same spelling is read differently by different people.

Case	What happens	What to do
Some names cannot be split, or are split wrongly	The built-in dictionary matches over three hundred family names by longest match, but rare names, compound names and names derived from places fall outside it. More fundamentally, the family-name-plus-given-name frame is not universal: Iceland has no shared family names at all, using patronymics instead, and some people in Indonesia have a single name. Even within Japan there are people with a middle name, people who use a working name alongside a legal one, and people who carry a maiden name alongside a married one. Getting the split wrong is not merely a data error; it means getting a person name wrong, which is a discourtesy.	Have a person check every row marked low confidence. The reason this tool reports a confidence at all is that reviewing every row by eye is impractical, so it narrows down which rows to look at. A name split on whitespace is high, a dictionary longest-match is medium, and everything else is low, so checking only the low rows gives you the most correction per unit of effort. More important as a design question is whether the split is needed at all: a form with separate family and given name fields turns away users outside Japan and anyone with a single name. If all you want is to address someone by name, no split is required — take the whole name in one field and display it as given. The errors introduced by splitting usually outweigh the inconvenience of not splitting.
The reading is blank, or wrong	One spelling can have several readings, and only the person concerned knows which is right. The surname written `東海林` is read both Shoji and Tokairin, and both are real families. `新谷` is Shintani, Niiya or Araya; `中田` is Nakata or Nakada — and these are not regional variants but differ from household to household, so no amount of dictionary coverage settles them mechanically. Given names are freer still: which reading attaches to a set of characters is entirely the choice of whoever named the child. For unusual or old-form characters, the dictionary simply has no entry and the reading comes back blank.	The only correct answer is to have the person supply the reading. On an application or a signup form, put a reading field next to the name field — not relying on estimation is best both for accuracy and as a matter of courtesy. This tool earns its place filling in existing data that accumulated without a reading field, and even then do not store the estimate as a settled value: keep it in a form that shows it is an estimate — a separate column, a flag — and overwrite it when you have an opportunity to confirm with the person. Use an estimated reading directly for sorting a roster or for a text-to-speech address label and the error becomes visible to everyone.
Using the high-accuracy mode without checking what it implies	The high-accuracy mode runs MeCab morphological analysis on the server, which means the names you entered travel over the network. This site keeps no logs and discards the input immediately after processing, but the fact of transmission is itself a constraint in some settings: customer lists, employee records and patient information are commonly covered by organisational rules forbidding transmission to external services. A name can identify a person on its own, so switching the mode purely because it is more accurate may put you in breach of those rules.	Try the default in-browser mode first and see whether it is enough. A name already separated by a space needs neither dictionary nor analysis and comes back as high confidence. In practice most roster data is already space-separated, so the high-accuracy mode is only called for when you have a body of names with no space in them. If you do need it, check your organisational rules before using it — the test is whether you would be allowed to email that file outside the organisation. If not, installing MeCab locally and processing it in your own environment is the correct choice (`brew install mecab mecab-ipadic` or `apt install mecab mecab-ipadic-utf8`): the same dictionary and the same engine, so the results match what you would get here.

A name is personal data. The default mode on this page runs entirely in your browser and transmits nothing, but from the moment you download the result as CSV or TSV, looking after it is your responsibility: a customer list living on in the downloads folder, or being shared under a filename that still contains the project name, are more likely incidents than anything in the processing itself. Decide how you will handle it — no project or company name in the filename, delete it when the work is done — before you start. Finally, one point for anyone in a position to shape the design: storing a name split into two fields assumes more than most systems actually need. If it is only for addressing someone, no split is required; if it is for sorting, what you need is the reading, not the split. Count how many places genuinely need the family name alone — addressing someone formally, grouping by surname — and you will see whether that is worth staking the accuracy of every record on. Consider not splitting at all as the first option.

📖 How to Use

1

Paste full names

One full name per line. Works with or without whitespace.
2

Run split

Click Split. Strategies: whitespace, 300+ surname dictionary, 2-char fallback.
3

Download CSV / TSV

Low-confidence rows can be edited inline before exporting as CSV or TSV.

❓ Frequently Asked Questions

How is confidence determined?

Whitespace split gives High. The 300+ surname dictionary longest-match gives Medium. Everything else (2-character fallback) gives Low. Readings come from the surname, given-name, compound, and single-kanji dictionaries via longest match.

Are compound surnames handled?

The 300+ dictionary includes major compound surnames (Sasaki, Hasegawa, Igarashi, etc.) and uses longest-match. Rare names may still be misclassified.

What about foreign or katakana-only names?

Not supported. Add a space manually or correct the result before export.

🔗 Related Tools

🐛 Found a bug or issue with this tool?

Free to use, no signup. Even just the steps to reproduce are helpful. Reports go directly to the operator and help us fix issues.