n2words
Numbers to words. 50 languages, 72 regional variants. Zero dependencies.
Why n2words?
- Pure Functions — Each language exports standalone functions. No classes, no configuration, no side effects.
- Tree-Shakeable — Import only what you need. Unused exports are eliminated by modern bundlers.
- Tiny Bundles — ~2.5-3.5 KB gzipped per language with all three forms, less per form. No bloat.
- Multiple Forms — Cardinal ("forty-two"), ordinal ("forty-second"), and currency ("forty-two dollars")
- 50 Languages, 72 Regional Variants — European, Asian, Middle Eastern, African, and more — most importable via a single bare-tag code (
n2words/de), no region required - Zero Dependencies — Works everywhere: Node.js, browsers, Deno, Bun
- BigInt Support — Accepts
bigint(and numeric-string) input, so large values keep full precision - Type-Safe — Full TypeScript support with generated
.d.tsdeclarations
Quick Start
npm install n2words
import { toCardinal } from 'n2words/en'
import { toCardinal as es } from 'n2words/es'
toCardinal(42) // 'forty-two'
es(42) // 'cuarenta y dos'
Forms
n2words converts numbers to words in multiple forms:
| Form | Function | Example |
|---|---|---|
| Cardinal | toCardinal(42) |
"forty-two" |
| Ordinal | toOrdinal(42) |
"forty-second" |
| Currency | toCurrency(42.50, { currency }) |
"forty-two dollars and fifty cents" |
import { toCardinal, toOrdinal, toCurrency } from 'n2words/en'
toCardinal(1234) // 'one thousand two hundred thirty-four'
toOrdinal(1234) // 'one thousand two hundred thirty-fourth'
toCurrency(1234.56, { currency: 'USD' }) // 'one thousand two hundred thirty-four dollars and fifty-six cents'
Upgrading from v5? See docs/migration-v6.md — three breaking changes, all in
toCurrency.
toCurrency is the one form a bare tag won't guess at. en names a language,
and a default currency belongs to a country, so n2words/en requires the
currency to be named. Import the region-qualified code when you want its
default:
import { toCurrency } from 'n2words/en'
import { toCurrency as usd } from 'n2words/en-US'
toCurrency(42.50) // throws TypeError — en is a language, not a locale
toCurrency(42.50, { currency: 'GBP' }) // 'forty-two pounds and fifty pence'
usd(42.50) // 'forty-two dollars and fifty cents'
A currency-supporting language can also name an amount in a currency other than its own
default, via the currency option — the set of currencies it knows is validated, so an
unsupported code throws rather than silently guessing. pt-BR is one of the few languages
with no bare-tag entry point (Brazilian and European Portuguese use different numbering
systems, not just different currencies — see LANGUAGES.md), so it's
imported by its full code:
import { toCurrency } from 'n2words/pt-BR'
toCurrency(42.50) // 'quarenta e dois reais e cinquenta centavos' (BRL, the default)
toCurrency(42.50, { currency: 'EUR' }) // 'quarenta e dois euros e cinquenta centavos'
toCurrency(42.50, { currency: 'CAD' }) // throws RangeError — pt-BR doesn't know CAD (yet)
You don't have to hardcode that set or catch the RangeError to discover it — every
currency-supporting language exports it, so it's readable at runtime and narrowed to a
literal union in TypeScript:
import { currencyDefaults, currencyValues } from 'n2words/pt-BR'
currencyDefaults.currency // 'BRL'
currencyValues.currency // ['BRL', 'USD', 'EUR', 'GBP', 'JPY']
This generalizes: every form that takes options exports its defaults
(cardinalDefaults, ordinalDefaults, currencyDefaults), and any option with a fixed set
of allowed values — currency, gender — also exports that set as <form>Values. Boolean
and free-string options have no such set, so they have no Values entry.
LANGUAGES.md lists every option, type and default per
language.
Each language implements one or more of these forms — see LANGUAGES.md for per-language coverage.
Range: each form spells values up to the largest scale word it knows, then throws a
RangeErrorrather than inventing vocabulary — and the ceiling varies by language and form (e.g.es-EScardinals reach 10^30 − 1, ordinals only 10^9 − 1). Cardinal and currency accept negatives and decimals; ordinal is positive integers only.
Usage
ESM (Node.js, modern bundlers):
import { toCardinal } from 'n2words/de' // Single form, bare-tag entry point
import { toCardinal, toOrdinal } from 'n2words/de' // Multiple forms
import { toCardinal as fr } from 'n2words/fr' // Aliased import
Most languages are imported this way — a bare BCP 47 primary subtag
(n2words/de, n2words/en, n2words/fr, ...), no region required. Each bare
tag is a documented alias for one specific regional/script variant (see
LANGUAGES.md's "Languages" table for the full list and each
one's default). Import the full region/script-qualified code instead when you
need a specific variant (n2words/en-GB, n2words/fr-BE) or when the
language has no single safe default — a handful of languages whose variants
genuinely diverge in script or core numbering grammar (Chinese, Portuguese,
Serbian, Amharic) require the full code:
import { toCardinal } from 'n2words/zh-Hans-CN'
A bare tag forwards toCardinal and toOrdinal untouched — they're the same
functions its target exports. The one difference is toCurrency, which
requires an explicit currency because a language tag carries no country to
take a default from (see
docs/bare-tag-aliases.md).
Browser (CDN):
<!-- ESM (recommended) -->
<script type="module">
import { toCardinal } from 'https://cdn.jsdelivr.net/npm/n2words/dist/en.js'
console.log(toCardinal(42)) // 'forty-two'
</script>
<!-- UMD (legacy script tags) -->
<script src="https://cdn.jsdelivr.net/npm/n2words/dist/en.umd.js"></script>
<script>
n2words.en(42) // 'forty-two'
n2words.ordinal.en(42) // 'forty-second'
n2words.currency.en(42.50, { currency: 'USD' }) // 'forty-two dollars and fifty cents'
// `en` is a language, so currency must be named — or load dist/en-US.umd.js
// and call n2words.currency.enUS(42.50) for a locale's own default.
</script>
dist/{code}.js carries all three forms. A page that needs only one can fetch
just that form from dist/{code}/{form}.js — cardinal, ordinal or
currency:
<script type="module">
import { toCurrency } from 'https://cdn.jsdelivr.net/npm/n2words/dist/en-US/currency.js'
console.log(toCurrency(42.50)) // 'forty-two dollars and fifty cents'
</script>
Under half the bytes for cardinal or ordinal, since the other two forms and
everything only they reach are never in the file. Currency is the heaviest of
the three: it carries the shared currency-name matrix, so en-US can quote any
of the currencies English has words for, not just USD.
en-US |
bytes |
|---|---|
dist/en-US.js (all three forms) |
9,227 |
dist/en-US/cardinal.js |
4,293 |
dist/en-US/ordinal.js |
3,815 |
dist/en-US/currency.js |
5,258 |
This is for CDN consumers only. If you install from npm and use a bundler,
import { toCurrency } from 'n2words/en-US' already drops the forms you don't
import — the package is sideEffects-free and each form is an independent
export, so there is nothing to opt into.
See LANGUAGES.md for all language codes and available forms.
Supported Languages
See LANGUAGES.md for the complete list with codes and options.
Highlights: Arabic, Chinese (Simplified/Traditional), English, French, German, Hindi, Japanese, Korean, Portuguese, Russian, Spanish, and many more.
Compatibility
- Node.js: 22+
- Browsers: Chrome 67+, Firefox 68+, Safari 14+, Edge 79+ — any browser with BigInt support
- Runtimes: Deno, Bun, Cloudflare Workers
Requires BigInt support (cannot be polyfilled).
Performance
n2words is optimized for both size and speed:
- ~2.5-3.5 KB gzipped per language, all three forms included
- Individual language imports enable tree-shaking
- No runtime dependencies
- BigInt modulo operations (no string manipulation)
- Pure functions with no shared state
- Minimal memory allocation per conversion
Run npm run bench to measure on your hardware.
Contributing
We welcome contributions! Add a new language or improve existing ones:
npm run lang:add -- <code> # Scaffold a new language (BCP 47 code)
npm test # Run full test suite
Also welcome: bug reports, feature requests, and documentation improvements.
- Contributing Guide — How to contribute and add languages
- Code of Conduct — Community standards
License
MIT Wael TELLAT, Tyler Vigario & contributors