jschardet

jschardet is a character encoding detector for JavaScript. Runs in Node.js and browsers with zero runtime dependencies.

jschardet 4 is a ground-up TypeScript port of chardet 7. It's much faster and more accurate than jschardet 3 and keeps its API; see Upgrading from jschardet 3 for what behaves differently.

The API is detect() and detectAll(), returning encoding, confidence, language, and mimeType.

Features

99.4% accuracy on 3,138 test files, up from 43.1% in jschardet 3, with ~6× the throughput and ~9× lower peak memory. Language detection for every result. MIME type detection for binary files.

	jschardet 4.0.0	jschardet 3.1.4	chardet 7.7.0 (Python)
Accuracy (3,138 files)	99.4%	43.1%	99.5%
Speed	777 files/s	122 files/s	234 files/s
Language detection	94.5%	—	94.5%
Peak memory	90.8 MiB	825.9 MiB	37.3 MiB
Bundle size (min / gzip)	1,149 / 695 KiB	334 / 119 KiB	—
Cold start (import + first detect)	105.4 ms	26.3 ms	138.5 ms
Runs in browsers	yes	yes	—
MIME type detection	yes	no	yes
License	0BSD	LGPL	0BSD

Compared to jschardet 3, v4 has a larger bundle and a ~105 ms first-call cost. Both come from shipping a larger detection model; the model decompresses once on first detect() and stays in memory afterwards, so subsequent calls run at full speed.

See docs/performance.md for the full benchmark methodology and per-encoding accuracy.

Installation

npm

npm install jschardet

Both module systems are supported:

import { detect } from 'jschardet';         // ESM
const { detect } = require('jschardet');    // CommonJS

Browser

Copy and include jschardet.min.js in your page (attaches a global jschardet). For ESM, use jschardet.esm.min.js instead. Unminified builds and source maps are in dist/.

The library is also available via jsDelivr:

Format	URL
IIFE (<script src>)	https://cdn.jsdelivr.net/npm/jschardet
ESM (<script type="module">)	https://cdn.jsdelivr.net/npm/jschardet/dist/jschardet.esm.min.js

Classic script tag (after copying jschardet.min.js next to your HTML):

<script src="jschardet.min.js"></script>
<script>
  console.log(jschardet.detect("\xc3\xa0\xc3\xad\xc3\xa0\xc3\xa7\xc3\xa3"));
</script>

ESM (after copying jschardet.esm.min.js):

<script type="module">
  import { detect } from './jschardet.esm.min.js';
  console.log(detect("\xc3\xa0\xc3\xad\xc3\xa0\xc3\xa7\xc3\xa3"));
</script>

Quick start

import { detect, detectAll } from 'jschardet';

// string — ASCII
detect("Python is a great programming language for beginners and experts alike.")
// { encoding: 'ascii', confidence: 1, language: 'en', mimeType: 'text/plain' }

// Uint8Array — "The naïve approach doesn't always work in complex systems." in UTF-8
detect(new TextEncoder().encode("The naïve approach doesn't always work in complex systems."))
// { encoding: 'utf-8', confidence: 0.84, language: 'en', mimeType: 'text/plain' }

// Uint8Array — "日本語の文字コード検出テストです。" in EUC-JP
detect(new Uint8Array([
  0xc6, 0xfc, 0xcb, 0xdc, 0xb8, 0xec, 0xa4, 0xce, 0xca, 0xb8,
  0xbb, 0xfa, 0xa5, 0xb3, 0xa1, 0xbc, 0xa5, 0xc9, 0xb8, 0xa1,
  0xbd, 0xd0, 0xa5, 0xc6, 0xa5, 0xb9, 0xa5, 0xc8, 0xa4, 0xc7,
  0xa4, 0xb9, 0xa1, 0xa3,
]))
// { encoding: 'EUC-JP', confidence: 0.56, language: 'ja', mimeType: 'text/plain' }

// Buffer (Node.js) — "Le café est une boisson très populaire en France et dans le monde entier." in windows-1252
const results = detectAll(Buffer.from([
   76, 101,  32,  99,  97, 102, 233,  32, 101, 115, 116,  32, 117, 110, 101,
   32,  98, 111, 105, 115, 115, 111, 110,  32, 116, 114, 232, 115,  32, 112,
  111, 112, 117, 108,  97, 105, 114, 101,  32, 101, 110,  32,  70, 114,  97,
  110,  99, 101,  32, 101, 116,  32, 100,  97, 110, 115,  32, 108, 101,  32,
  109, 111, 110, 100, 101,  32, 101, 110, 116, 105, 101, 114,  46,
]))
for (const r of results.slice(0, 4)) {
  console.log(r.encoding, r.confidence.toFixed(2));
}
// Windows-1252 0.32
// iso8859-15 0.32
// ISO-8859-1 0.32
// MacRoman 0.31

API

detect(buffer, options?)

Accepts a string, Uint8Array, ArrayBuffer, or any ArrayBufferView (Node Buffer and DataView work too). Returns the best match as an object:

Field	Type	Description
encoding	string | null	Detected encoding name, or null if unknown
confidence	number	Score from 0.0 to 1.0
language	string | null	Language hint when available
mimeType	string | null	MIME type hint when available

detectAll(buffer, options?)

Same input types as detect. Returns all candidates above the confidence threshold (default 0.20), sorted by confidence. At least one result is always returned when the built-in threshold applies.

Options

interface IOptionsMap {
  minimumThreshold?: number;      // override default 0.20 for detectAll filtering
  detectEncodings?: string[];     // allowlist of encoding names to consider
  excludeEncodings?: string[];    // blocklist of encoding names
}

enableDebug()

Logs full candidate lists to the console from detect / detectAll (useful when tuning thresholds or allowlists).

CLI

jschardet somefile.txt
# somefile.txt: utf-8 with confidence 1

jschardet --minimal somefile.txt
# utf-8

# Include detected language
jschardet -l somefile.txt
# somefile.txt: utf-8 en (English) with confidence 1

# Only consider specific encodings
jschardet -i utf-8,windows-1252 somefile.txt
# somefile.txt: utf-8 with confidence 1

# Pipe from stdin
cat somefile.txt | jschardet
# stdin: utf-8 with confidence 1

Supported encodings

Same encodings as chardet (aliases and encoding-era filters are documented there).

Modern Web

ascii, big5hkscs, cp874, cp932, cp949, euc-jis-2004, euc-kr, gb18030, hz-gb-2312, iso-2022-kr, iso2022-jp-2, iso2022-jp-2004, iso2022-jp-ext, koi8-r, koi8-u, shift_jis_2004, tis-620, utf-16, utf-16-be, utf-16-le, utf-32, utf-32-be, utf-32-le, utf-7, utf-8, utf-8-sig, windows-1250, windows-1251, windows-1252, windows-1253, windows-1254, windows-1255, windows-1256, windows-1257, windows-1258

Legacy ISO

iso-8859-1, iso-8859-2, iso-8859-3, iso-8859-4, iso-8859-5, iso-8859-6, iso-8859-7, iso-8859-8, iso-8859-9, iso-8859-10, iso-8859-13, iso-8859-14, iso-8859-15, iso-8859-16, johab

Legacy Mac

mac-cyrillic, mac-greek, mac-iceland, mac-latin2, mac-roman, mac-turkish

Legacy Regional

cp1006, cp1125, cp720, hp-roman8, koi8-t, kz-1048, ptcp154

DOS

cp437, cp737, cp775, cp850, cp852, cp855, cp856, cp857, cp858, cp860, cp861, cp862, cp863, cp864, cp865, cp866, cp869

Mainframe (EBCDIC)

cp1026, cp1140, cp273, cp424, cp500, cp875

chardet module

The upstream chardet API is available as-is via the chardet named export. Use UniversalDetector for streaming detection over large files or network streams:

import { chardet } from 'jschardet';
import { createReadStream } from 'node:fs';

const detector = new chardet.UniversalDetector();
for await (const chunk of createReadStream('unknown.txt')) {
  detector.feed(chunk);
  if (detector.done) break;
}
console.log(detector.close());
// { encoding: 'utf-8', confidence: 1, language: 'en', mimeType: 'text/plain' }

Upgrading from jschardet 3

detect(), detectAll(), enableDebug(), and the minimumThreshold and detectEncodings options keep their v3 signatures, and a string is still read as one byte per character. What changes:

• Results. Many inputs detect differently; that is where the accuracy gain comes from. Encoding names follow chardet's spelling, e.g. utf-8 where v3 returned UTF-8, so compare names case-insensitively.
• detectEncodings accepts v3 names except four encodings v4 does not detect, which throw: EUC-TW, ISO-2022-CN, X-ISO-10646-UCS-4-3412, and X-ISO-10646-UCS-4-2143. When the list leaves out windows-1252 and no listed encoding matches, the result is null and a warning is logged with console.warn.
• Confidence scale. Correct answers often score far lower than in v3: the Big5 sample from v3's README scored 0.99 in v3 and scores 0.08 in v4. Re-tune any threshold chosen against v3 numbers.
• detect() always returns its best guess. v3 returned { encoding: null, confidence: 0 } when the best guess scored below minimumThreshold (default 0.20); v4's detect() ignores minimumThreshold, so check confidence yourself. minimumThreshold filters detectAll() instead, which v3 did not filter.
• Empty input returns utf-8 with confidence 0.1, where v3 returned { encoding: null, confidence: 0 }.
• Result type. Results gain language and mimeType, and encoding is typed string | null (v3: string), so TypeScript code under strictNullChecks needs a null check.
• UniversalDetector, exported but undocumented in v3, is now chardet.UniversalDetector (see chardet module). It has chardet's API: feed() takes a Uint8Array rather than a string, and the constructor takes chardet's options (includeEncodings rather than detectEncodings, no minimumThreshold).

License

0BSD, same as chardet.