TypographizerJS

Typographizer automates everything some things you don’t want (or need) to know about typography.

---

Acknowledgement

The JS version is inspired by the Typographizer by Frank Rausch, written in Swift. 🙇‍

Usage

Here be dragons 🐉

This package uses RegExp lookbehinds. These are supported in evergreen browsers and Node ≥ 16 (out of the box).

You can use this in the browser, but know what you’re signing up for: formatting and reinserting long strings costs time. Measured numbers live in the Performance section below.

In general, this is more of a fun project. So, use with caution.

Installation

npm install typographizer-js

Import the bundle into your project:

// ESM
import TypographizerJS from 'typographizer-js'

// CommonJS
const TypographizerJS = require('typographizer-js')

TypeScript types ship with the package.

And crate a new instance:

const Typographizer = new TypographizerJS()

Note: typographize() is safe to use on Markdown documents and rendered HTML. Inline HTML tags (a, strong, em) are preserved, and code samples are skipped: fenced code blocks (``` and ~~~), inline code spans (…), as well as `

` and `` elements are left untouched. Everything outside of code segments gets the usual fixes. When code segments are present, whitespace is neither trimmed nor collapsed, so the document structure stays intact.

Once you have your Typographizer instance available you can format strings using the typographize() method.
typographize returns a Promise. So you need to either await the return value or use .then().
const str = `I introduced myself, "Hi, I'm Oscar".`

/**
 * Using await
 */
const formatted = await Typographizer.typographize(str)

console.log(formatted)
// I introduced myself, “Hi, I’m Oscar“.

/**
 * Using .then()
 */
Typographizer.typographize(str)
  .then((str) => {
    console.log(str)
    // I introduced myself, “Hi, I’m Oscar“.
  })

Typographizer uses a bunch of different methods under the hood. All of them can be called on their own and skip code segments as well — with the exception of trimWhitespace(), which operates on the whole string by definition. For details, take a look at the API section below.
Options
By default quotes will be formatted to match en_US style. If you need different formatting, pass a language code while initializing.
const Typographizer = new TypographizerJS('fr')

You can find a list of supported languages further down this file.
⚠️ To avoid side effects and confusion, It is not possible to change these options after you created your instance. If you want to format text in multiple languages, you need to create multiple instances.
API
All these methods return Promises. Usage instructions for typographize usage apply.
Methods



Method
Description




fixApostroph(str)
Replace ', ´ and ` used as an apostroph


fixEllipsis(str)
Formats all occurences of ... to use the correct … character


fixAiryPunctuation(str)
Finds punctuation that is surrounded by whitespace


formatQuotes(str)
Format quotes, but not whitespace or apostrophes


formatOpeningQuotes(str)
Format all opening quotes


formatClosingQuotes(str)
Format all closing quotes


trimWhitespace(str)
Remove whitespace from the beginning and end of a string, replaces multiple spaces with a single one


typographize(str)
Takes a string and applies all optimizations



Performance
The benchmark measures

the full typographize() pipeline for English and non-English languages (en_US, de, fr, pl) at 256 B, 1 KB, 10 KB, 100 KB and 1 MB, and
every public method individually at 10 KB, 100 KB and 1 MB.

Each measurement is the best of 27 runs after 5 warmups: GC pauses and scheduler noise only ever add upward error, so the minimum is the most reproducible estimate of actual cost.
Every pull request is gated by an A/B comparison against main: the benchmark runs both checkouts back-to-back on the same machine, interleaved per suite so both sides share the same noise window, and fails when any suite is more than 1.5× slower than main. Hardware and load differences cancel out this way — absolute thresholds were tried first and proved unreliable: they break across machines, across CI runners (whose timing varies up to 4× run to run) and on your own machine the moment something else eats CPU.
Latest numbers (2026-10-26, Node v26.12.1, Apple M5, macOS 26 — absolute values vary with the machine, the 1.5× ratio is what gets enforced):
Full pipeline (typographize()), best run in ms:



Size
en_US
de
fr
pl




256 B
0.00
0.00
0.00
0.01


1 KB
0.01
0.01
0.00
0.01


10 KB
0.04
0.04
0.04
0.04


100 KB
0.35
0.36
0.30
0.36


1 MB
3.56
3.60
3.01
3.59



Single methods (en_US), best run in ms:



Method
10 KB
100 KB
1 MB




trimWhitespace
0.00
0.02
0.24


formatQuotes
0.02
0.20
2.02


formatOpeningQuotes
0.01
0.10
0.95


formatClosingQuotes
0.01
0.09
0.91


fixApostroph
0.00
0.02
0.20


fixEllipsis
0.00
0.01
0.13


fixAiryPunctuation
0.01
0.07
0.69



To run the benchmark yourself:
npm run bench             # print absolute numbers (reporting only)
npm run bench:compare     # A/B against main; creates a temporary worktree
npm run bench:compare -- path/to/other-checkout  # A/B against any checkout

Supported languages
Language support has been ported from the original Typographizer.
Supported languages are:



Language
Shortcode




Bosnian
bs


Czech
cz


Danish
da


Dutch, Flemish
nl


British English
en_GB


American English (default)
en_US


Estonian
es


Finnish
fi


French
fr


German
de


German (Suisse)
de_CH


German (Liechtenstein)
de_LI


Hungarian
hu


Icelandic
is


Lithuanian
lt


Latvian
lv


Polish
pl


Romanian
ro


Slovak
sk


Slovenian
sl


Swedish
sw



This list is missing Japanese and Hebrew, which are supported by Typographizer. Hebrew is missing due to complicated support. Japanese requires extended Unicode checks.