Turn words into flowing cursive script and calligraphy lettering with one click.
When you paste a cursive result from this page, you have not applied a font. You have swapped every letter for a different character with its own code point, and that character carries its swirl wherever it goes because the shape belongs to its identity rather than to a font file. All 34 rows in the Cursive & script set draw on the Mathematical Alphanumeric Symbols block at U+1D400 to U+1D7FF, added in Unicode 3.1 in 2001, the first release to use characters beyond the Basic Multilingual Plane. The standard describes that block as notation for mathematics, where the difference between an italic variable and a script one carries meaning a reader has to recover from the plain text itself.
That origin explains nearly every behaviour you will run into. A script capital that looks like a decorated A is not an A; compare the two strings in any programming language and they differ. Type your cursive name into a search box and the plain spelling will not find it. Sort a column of them and they will not sit beside their plain twins the way you expect. None of that is a bug in the generator. It follows directly from a styled letter being a separate character, which is also why the effect survives being copied between apps that share no formatting at all.
The block gives you two script series, and they are not equivalent. Mathematical Script begins at U+1D49C with the script capital A and runs through the lowercase z. Mathematical Bold Script begins at U+1D4D0, ends at U+1D503, and carries a heavier stroke. Pick either row and the transformation is a straight per-letter lookup: A becomes the script A, b becomes the script b, and anything without a counterpart passes through untouched.
The difference that matters is completeness. The bold series holds all fifty-two letters in one unbroken run. The lighter series does not. Positions inside it are reserved and hold no character at all, because those letters had already been encoded years earlier in the Letterlike Symbols block at U+2100 to U+214F, and Unicode does not encode the same character twice. Eight capitals are missing from the run, B, E, F, H, I, L, M and R, along with three lowercase, e, g and o.
If a cursive word has ever had one letter sitting oddly against the rest, this is usually why. A generator that simply adds a fixed offset to each letter's code point walks straight into those reserved slots and emits an undefined character or an empty box. Every style here maps letter by letter instead, so the gaps are filled deliberately rather than by arithmetic.
Look at which letters are missing: B, E, F, H, I, L, M, R, and the lowercase e, g and o. Those are not obscure. They are the letters people put first in a monogram and first in a name, so the gap lands precisely where it is most visible. There are two ways to fill it and neither is free.
The first is to reach into Letterlike Symbols for the script characters already sitting there: U+212C for the script capital B, U+2130 and U+2131 for E and F, U+210B, U+2110, U+2112 and U+211B for H, I, L and R, U+2133 for M, and U+212F, U+210A and U+2134 for the small e, g and o. You get genuinely looped shapes. The cost is that these live in the Basic Multilingual Plane while the rest of the alphabet does not, so they usually resolve from a different font file and can appear heavier, wider or on a slightly different baseline than their neighbours.
The second is what the lighter script row here does: it borrows the mathematical italic letters instead, so those eleven come out sloped rather than looped. The word stays visually even because every glyph is drawn from one font, and the trade is that eleven letters in it are not script at all.
Real cursive is defined by the join. The pen does not leave the paper, so the exit stroke of one letter becomes the entry stroke of the next, and a letter changes shape depending on what sits either side of it. Unicode does support that behaviour where a writing system requires it: Arabic letters carry a joining type and take initial, medial, final and isolated forms. Latin letters carry none of it. The script characters in the maths block are non-joining, and no property in the standard tells a renderer to bend one into the next.
So what you get is a run of separate, self-contained ornamental glyphs set side by side with ordinary letter spacing. Each was drawn to stand alone in an equation, where a script capital is a single symbol naming a quantity rather than part of a word. Strung together they read as slanted, looping italic letters that never quite touch, which is exactly the effect people describe as looking slightly off without being able to say why. The connecting strokes were never encoded.
If joined handwriting is what you actually need, only one thing produces it: a real typeface with connections drawn into the glyphs and the pairs kerned, applied to plain letters. That is a font choice, not a character substitution.
The word cursive covers traditions a calligrapher would never confuse. Copperplate, or English roundhand, is the engraved hand of the eighteenth-century writing masters: extreme contrast between hairline and swell, and a steep constant slant. Spencerian is the American business hand taught from the mid-nineteenth century by Platt Rogers Spencer, lighter, faster, with flourished capitals. Brush script is a twentieth-century advertising style with an even marker-like stroke and a casual bounce.
Unicode encodes none of that distinction. There is one script alphabet and one bold script alphabet, and which tradition they resemble is decided entirely by the font your reader's device happens to use. The same string can look like engraved roundhand on one machine and like a soft rounded chancery on another, because you specified a character, not a design.
What this page varies is packaging rather than penmanship. Of the 34 rows in the set, two are those script alphabets and thirty-two wrap the bold script in decorative frames, wings, crowns and brackets that change the silhouette of a short name without changing a single letter. Judge a row by how it looks at your actual size, not by its label.
Cursive works when the string is short, decorative, and read by a person rather than a machine. A display name on a profile, a heading inside a graphic, a divider in a newsletter: cases where the whole point is that the words look unlike everything around them, and where nobody needs to search, quote or hear them.
Budget the length before you commit. A character from the maths block costs two UTF-16 code units, and platforms that count that way charge you double. X weights characters outside a small set of Latin and punctuation ranges at two, so a post capped at 280 characters holds roughly 140 script letters. The framed rows compound it, since every wing, crown and bracket is another character on each side of the name, which is fine inside Instagram's 30-character name field if you count first and ruinous if you paste and hope.
Then check it on the device your reader will use, not the one you are typing on. Coverage for these ranges depends on which fonts are installed, and a phone, a television app and a desktop browser can each resolve them from a different file. If a name will be seen by thousands of people, look at it on two operating systems before you save it.
Anything longer than a line is the wrong place for these. A paragraph of script characters is slow going even when every glyph renders, because the letterforms are ornamental, closely similar, and never received the spacing work a real text face gets.
Search is the next casualty. Browser find, site search and hashtag matching all compare characters, so a cursive spelling is invisible to anyone typing the plain one, and hashtags built from these letters usually match nothing. Some systems go the other way and normalise them out of existence: these characters carry compatibility decompositions, so any pipeline applying NFKC turns your script back into ordinary ASCII. That is why a styled name can survive happily in a bio and collapse the moment it is used as an identifier.
Assistive technology is the cost people forget. Screen readers handle this block inconsistently: some announce nothing, some read a character name for every letter, some spell the word out. A visitor using one hears noise or silence where your name should be, and unlike a font choice there is no plain layer underneath to recover. Where the text carries information rather than mood, a form label, a warning, instructions, a link, use ordinary letters and put the decoration somewhere that does not have to be understood.
Two jobs bring most people to a cursive generator, and they have different answers.
For a wedding invitation, an order of service or anything that gets printed, do not use these characters. Print hands you control of the typeface, which is the entire advantage, and a drawn script offers joined strokes, ligatures and swashes that no character substitution can reproduce. Unicode script is for the places that refuse to let you pick a font: the display name on an RSVP page, a shared album title, a messaging app with no formatting.
For a signature, meaning the sign-off at the end of an email or a name under a note, the copy button here has a better route than pasting styled characters. Copy as rich text and the page hands the destination your plain letters plus a real script font stack, so Word, Google Docs and Gmail show genuine cursive that stays searchable, spell-checkable and readable aloud. Save as .docx and the same substitution is written into the file. The styled-character version is the one to reach for only when the destination strips formatting.
Be clear about what styling does not do. Typing your name in looped characters does not make it a stronger signature; what gives an electronic signature its weight is the process around it, never the shape of the letters.
Because script digits do not exist. The block carries styled digit sets for bold, double-struck, sans-serif and monospace, but no script series and no script punctuation, so a generator has nothing to map 0 to 9 onto and passes them through untouched. A dated line will always mix looped letters with ordinary figures; spell the number out if that bothers you.
Not from a normal keyboard layout, since nothing on it maps to characters beyond the Basic Multilingual Plane. In Microsoft Word you can type a hex code such as 1D4D0 and press Alt+X to convert it in place, and on macOS the Emoji and Symbols viewer lets you search by character name. Both are slower than copying a finished row.
It risks it. A crawler indexes the characters you actually published, so a title written in script characters may not match a query typed in plain letters, and no engine guarantees it will normalise them for you. Keep titles, headings and link text in ordinary characters, and put decoration in a short line that carries no information.
Two different mechanisms. Some platforms normalise or strip non-ASCII characters when text is re-posted or pushed through an API, which flattens the styling. Others keep the characters intact but render them on a device with different fonts installed, so the string simply looks different. Copy the quoted version and check whether the characters themselves survived.
None of the 34 here can be relied on, because the problem is the characters rather than the particular style. The accessible route on the web is the opposite approach: leave the letters plain and apply a script typeface with CSS, so the text underneath stays ordinary and the decoration is purely visual. In apps offering no styling, plain letters are the answer.
Modern file systems store Unicode and will usually accept them, but the tooling around your files is the risk. Command-line completion, older archive formats, sync clients, upload forms and any script assuming ASCII can mangle or reject the name. A file you have to rename before sending it anywhere costs more than the decoration is worth.
Because you never specified a design, only a set of characters. Each device picks whichever installed font covers the range, such as a maths-capable font on Windows or a system fallback on Android and iOS, and those differ in weight, slant and proportion. Mixed strings make it plainer still, since substituted letters can come from a different file.
No. There is no character or control code that joins Latin letters. The zero width joiner at U+200D affects emoji sequences and writing systems that already have joining behaviour, and it does nothing between Latin letters. Joined cursive comes only from a typeface whose glyphs are drawn with connections and whose spacing accounts for them.