Cursive Text Generator

Copy-and-paste cursive is not a font. Every letter is swapped for a different Unicode character from the Mathematical Alphanumeric Symbols block, which was added in Unicode 3.1 in 2001 so that equations could tell a script 𝒜 from a bold 𝐀 and a plain A. Because each one is its own character, the look survives being pasted into a bio or caption where you cannot choose a font.

Unicode only has two true script alphabets, light and bold. The other rows are italics: slanted letters rather than handwriting, but they are what many people mean by “cursive”, and they read more easily in long lines.

None of the styles has slanted or script digits, so numbers stay as they are unless you switch on bold digits in Fine-tune. Those are upright, so 𝓑𝓸𝓸𝓴 𝟐 can look a little stiff next to the script letters. Punctuation, emoji and letters outside A to Z, such as ß or ø, are also left plain. Accented letters are styled when the accent can be split off as a separate mark, covered in the FAQ below.

Eleven script letters were already in Unicode before the maths alphabets were added, in the Letterlike Symbols block from 1993, because physics and maths used them on their own. Unicode did not encode them twice. It left gaps in the new alphabet, which the standard itself calls holes, and the older characters fill them.

Fonts often draw the two blocks differently. A phone may take Letterlike Symbols from its main text font and the maths letters from a separate maths font, so ℯ can sit higher, look thinner or lean at another angle than the 𝓇 beside it.

Worked example. “Grace Hopper” in Script is 𝒢𝓇𝒶𝒸ℯ ℋℴ𝓅𝓅ℯ𝓇. Four of its eleven letters, both e’s, the H and the o, come from the older block, which is why the name can look uneven. In Bold script, 𝓖𝓻𝓪𝓬𝓮 𝓗𝓸𝓹𝓹𝓮𝓻, every letter comes from the same alphabet, because the bold set was new in 2001 and has no holes. If Script looks patchy on your screen, Bold script or Bold capitals is the fix. Italic has a single hole, the small h, which is ℎ, the Planck constant sign.

School cursive joins each letter to the next. These characters cannot. Each one is a separate symbol drawn by whatever font the app happens to use, and nothing in Unicode tells 𝓪 to reach across to 𝓫. Connected scripts such as Arabic join because fonts and shaping engines have rules for those letters; maths letters are designed to stand alone, the way variables do in an equation.

Joined-up lettering on screen comes from a script font applied to ordinary letters. The font swaps in connected shapes (OpenType contextual alternates) as it draws. That only works where you can pick a font: Word, Canva, a design app or a web page. In CSS there is even a generic cursive family, but it leaves the choice to the browser, which on Windows usually means Comic Sans MS, so name a real script font instead. For a page that looks written by hand, use convert text to handwriting, which draws your words as an image.

Script capitals come in two traditions. Chancery is the narrower, calligraphic hand with flat pen strokes; roundhand is the looped copperplate style of English writing masters. Mathematicians long treated them as the same letter, but some papers use both with different meanings, so Unicode 14.0 (2021) added a way to ask for one or the other: put the invisible variation selector U+FE00 after a script capital for chancery, or U+FE01 for roundhand. All 26 regular script capitals have both sequences; bold script and small letters have none.

The Script capitals menu in Fine-tune adds the selector for you. Most fonts ignore it, so in a typical app nothing changes; maths fonts are the likeliest to honour it. The selector stays in the text either way and counts towards limits: on X, ℋ︀ℴ𝓅ℯ with a chancery selector counts as 10 rather than 8. Leave the menu on “font default” unless you know your destination font supports the variants.

Maths letters live outside the Basic Multilingual Plane, so each one takes 4 bytes in UTF-8 and two UTF-16 units, which is what JavaScript’s length and many form limits count. The Letterlike holes take 3 bytes and one unit, so Script is slightly shorter than the other styles. X weighs characters differently: its counting rules give weight 1 only to a few ranges of common scripts and punctuation, and everything else, including all of these letters, counts as 2.

So a 160-character bio limit measured in UTF-16 units fits only about 80 cursive letters, and a 280-character post on X fits about 140. The status line under the text box shows the X count for the Script row as you type.

The Unicode Standard is blunt about this: characters from the Mathematical Alphanumeric Symbols block “should not be used to represent styling of nonmathematical text”. Nothing stops you, and for a display name or caption it is mostly harmless, but it is worth knowing what other software does with them.

• Search and matching. Every letter in all six styles has a compatibility mapping to its plain letter, so software that applies NFKC normalisation treats 𝓖𝓻𝓪𝓬𝓮 as Grace. Software that compares raw characters does not, and a search for “grace” can miss it entirely.
• Screen readers. Some read each letter by its Unicode name, such as “mathematical bold script capital G”, some spell it out, and some skip it. A cursive post can be unreadable to a blind follower, so keep names, links and anything important in plain letters.
• Missing fonts. Current phones and computers ship a font that covers these letters. Older devices, some smart TVs, game consoles and printers may show empty boxes, and Script may show boxes for some letters but not the eleven older ones.
• Usernames. Handles usually allow only plain letters, digits and a few marks, so cursive works in display names but not in the @name itself.

Paste cursive into the box and a Plain text row appears above the styles; there is no direction to switch or style to pick. It decodes any maths letter or digit, so bold, Fraktur 𝔉, double-struck 𝔻 and monospace text from other generators come back too, along with the Letterlike holes. It drops chancery and roundhand selectors, and leaves look-alikes that are not letters, such as ™ and ℃, alone.

Worked example. 𝒞𝒶𝒻ℯ́ 𝟐𝟎𝟐𝟔 is nine characters you can see, but the ℯ́ is two: ℯ (U+212F) followed by a combining acute accent (U+0301). Decoding turns the ℯ into e, recombines it with the accent and gives back “Café 2026” with a normal precomposed é, so it matches the original in search and spell-check.

In a spreadsheet or script you can do the same with NFKC normalisation, for example text.normalize('NFKC') in JavaScript or unicodedata.normalize('NFKC', text) in Python. Be aware that NFKC also rewrites other things, such as fi to fi, ² to 2 and ™ to TM, which this page does not touch.