Full-width ⇄ Half-width Converter
Convert 全角 (full-width) and 半角 (half-width) characters in both directions - Latin letters, numbers, symbols, spaces and katakana. Choose exactly which character types to touch, so you can normalise the numbers in a form without disturbing anything else, and use bulk mode to run the same change down a whole CSV column.
Why two widths exist at all
East Asian text is traditionally laid out on a square grid, one character per cell. Latin letters and digits are much narrower than a kanji, so a second, full-width set was encoded that occupies a whole cell and lines up with the surrounding text. Unicode keeps these in the Halfwidth and Fullwidth Forms block, laid out in the same order as ASCII - which is why converting them is exact arithmetic rather than a lookup table: full-width Ａ sits precisely 0xFEE0 code points above ASCII A.
Katakana works the other way round and is not simple arithmetic. Half-width katakana came from early systems that needed kana in a single byte, and it has fewer characters than the full-width set because voiced sounds are written as a base character plus a separate voiced mark. Converting ｶ plus a dakuten into ガ means combining two characters into one, and going back splits them apart again. This tool handles that composition properly rather than shifting code points and hoping.
The practical reason to care is data quality. Japanese and Korean web forms routinely receive phone numbers, postcodes and product codes in whichever width the user's input method happened to produce, so ０３-１２３４ and 03-1234 are the same number stored as completely different strings. Normalising width before comparing or storing is what makes search, deduplication and validation work.
Perguntas frequentes
What does 全角 半角 変換 (zenkaku hankaku henkan) mean?
It is Japanese for 'full-width half-width conversion'. 全角 (zenkaku) characters take up a full square cell, the same width as a kanji, while 半角 (hankaku) characters are the normal narrow ASCII forms. Converting between them is one of the most common text clean-up tasks in Japanese data entry, because the same phone number or product code can be typed either way.
What is the difference between full-width and half-width?
They are different Unicode characters that look like the same letter. Full-width Ａ is U+FF21 and half-width A is U+0041. To a person they read identically; to a computer they are no more equal than 'A' and 'B', which is why a search for a half-width phone number silently fails to match the full-width version stored in your database.
Does this work for Korean 전각/반각 text too?
Yes. Korean uses the same Unicode block for full-width (전각) and half-width (반각) Latin letters, digits and symbols, so converting Korean form data works exactly the same way. Hangul syllables themselves are always full-width and are never altered by this tool.
Will it change my kanji, hiragana or Hangul?
No. Only the character classes you tick are converted, and kanji, hiragana and Hangul syllables are never in scope. You can safely run a whole document through it to normalise just the numbers, or just the katakana, and everything else comes back byte for byte identical.
How does it handle ｶﾞ and ﾊﾟ - the voiced katakana?
Correctly, which is the part naive converters get wrong. In half-width katakana a voiced sound is two characters: the base kana plus a separate dakuten or handakuten mark. Full-width katakana has a single character for it. Converting ｶﾞ to full-width produces the one character ガ, and converting ガ back produces the two characters ｶﾞ again, so the round trip is lossless.
Can I convert just one column of a CSV?
Yes. Turn on bulk mode, switch it to CSV columns, and pick the column you want. Every other column is left exactly as it was, and the commas and quote marks that give the file its structure are never converted - so a full-width pass cannot turn your separators into full-width commas and break the file.
Is my text sent to a server?
No. The conversion is pure Unicode arithmetic and a small katakana table, both running in your browser. Nothing is uploaded, so it is safe to use on customer lists and other private data.