Charset Detection, for Everyone 👋

^{The Real First Universal Charset Detector}

^{Featured Packages}

^{In other language (unofficial port - by the community)}

A library that helps you read text from an unknown charset encoding.
Motivated by chardet, I'm trying to resolve the issue by taking a new approach. All IANA character set names for which the Python core library provides codecs are supported.

>>>>> 👉 Try Me Online Now, Then Adopt Me 👈 <<<<<

This project offers you an alternative to Universal Charset Encoding Detector, also known as Chardet.

Feature	Chardet	Charset Normalizer	cChardet
Fast	✅	✅	✅
Universal**	❌	✅	❌
Reliable without distinguishable standards	✅	✅	✅
Reliable with distinguishable standards	✅	✅	✅
License	Public Domain and/or LGPL-2.1*** restrictive	MIT	MPL-1.1 restrictive
Native Python	✅	✅	❌
Detect spoken language	✅	✅	N/A
UnicodeDecodeError Safety	❌	✅	❌
Whl Size (min)	500 kB	150 kB	~200 kB
Supported Encoding	99	99	40
Can register custom encoding	❌	✅	❌

** : They are clearly using specific code for a specific encoding even if covering most of used one.
*** : The vast majority of the code is issued from an LLM agent (Claude), even if the author label this project now as MIT in his own name, it's clearly debatable. Most jurisdictions on copyright laws would nullify the license. With my personal education, Public Domain or/and LGPL-2.1 is the most likely one based on Anthropic declarations about how they train their LLMs and the LGPL-2.1 itself (the original license as it's still the same statistical principle behind the scene, hugely refactored).

⚡ Performance

This package offer acceptable performances against Chardet. Here are some numbers.

Package	Accuracy	Mean per file (ms)	File per sec (est)
chardet 7	89 %	5 ms	200 file/sec
charset-normalizer	97 %	8 ms	125 file/sec

Package	99th percentile	95th percentile	50th percentile
chardet 7	32 ms	17 ms	1 ms
charset-normalizer	63 ms	29 ms	3 ms

updated as of Mars 2026 using CPython 3.12, and Chardet 7

Chardet's performance on larger file (1MB+) are very poor. Expect huge difference on large payload. No longer the case since Chardet 7.0+

Stats are generated using 400+ files using default parameters. More details on used files, see GHA workflows. And yes, these results might change at any time. The dataset can be updated to include more files. The actual delays heavily depends on your CPU capabilities. The factors should remain the same. Chardet claims on his documentation to have a greater accuracy than us based on the dataset they trained Chardet on(...) Well, it's normal, the opposite would have been worrying. Whereas charset-normalizer don't train on anything, our solution is based on a completely different algorithm, still heuristic through, it does not need weights across every encoding tables.

✨ Installation

Using pip:

pip install charset-normalizer -U

🚀 Basic Usage

CLI

This package comes with a CLI.

usage: normalizer [-h] [-v] [-a] [-n] [-m] [-r] [-f] [-t THRESHOLD]
                  file [file ...]

The Real First Universal Charset Detector. Discover originating encoding used
on text file. Normalize text to unicode.

positional arguments:
  files                 File(s) to be analysed

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         Display complementary information about file if any.
                        Stdout will contain logs about the detection process.
  -a, --with-alternative
                        Output complementary possibilities if any. Top-level
                        JSON WILL be a list.
  -n, --normalize       Permit to normalize input file. If not set, program
                        does not write anything.
  -m, --minimal         Only output the charset detected to STDOUT. Disabling
                        JSON output.
  -r, --replace         Replace file when trying to normalize it instead of
                        creating a new one.
  -f, --force           Replace file without asking if you are sure, use this
                        flag with caution.
  -t THRESHOLD, --threshold THRESHOLD
                        Define a custom maximum amount of chaos allowed in
                        decoded content. 0. <= chaos <= 1.
  --version             Show version information and exit.

normalizer ./data/sample.1.fr.srt

or

python -m charset_normalizer ./data/sample.1.fr.srt

🎉 Since version 1.4.0 the CLI produce easily usable stdout result in JSON format.

{
    "path": "/home/default/projects/charset_normalizer/data/sample.1.fr.srt",
    "encoding": "cp1252",
    "encoding_aliases": [
        "1252",
        "windows_1252"
    ],
    "alternative_encodings": [
        "cp1254",
        "cp1256",
        "cp1258",
        "iso8859_14",
        "iso8859_15",
        "iso8859_16",
        "iso8859_3",
        "iso8859_9",
        "latin_1",
        "mbcs"
    ],
    "language": "French",
    "alphabets": [
        "Basic Latin",
        "Latin-1 Supplement"
    ],
    "has_sig_or_bom": false,
    "chaos": 0.149,
    "coherence": 97.152,
    "unicode_path": null,
    "is_preferred": true
}

Python

Just print out normalized text

from charset_normalizer import from_path

results = from_path('./my_subtitle.srt')

print(str(results.best()))

Upgrade your code without effort

from charset_normalizer import detect

The above code will behave the same as chardet. We ensure that we offer the best (reasonable) BC result possible.

See the docs for advanced usage : readthedocs.io

😇 Why

When I started using Chardet, I noticed that it was not suited to my expectations, and I wanted to propose a reliable alternative using a completely different method. Also! I never back down on a good challenge!

I don't care about the originating charset encoding, because two different tables can produce two identical rendered string. What I want is to get readable text, the best I can.

In a way, I'm brute forcing text decoding. How cool is that ? 😎

Don't confuse package ftfy with charset-normalizer or chardet. ftfy goal is to repair Unicode string whereas charset-normalizer to convert raw file in unknown encoding to unicode.

🍰 How

• Discard all charset encoding table that could not fit the binary content.
• Measure noise, or the mess once opened (by chunks) with a corresponding charset encoding.
• Extract matches with the lowest mess detected.
• Additionally, we measure coherence / probe for a language.

Wait a minute, what is noise/mess and coherence according to YOU ?

Noise : I opened hundred of text files, written by humans, with the wrong encoding table. I observed, then I established some ground rules about what is obvious when it seems like a mess (aka. defining noise in rendered text). I know that my interpretation of what is noise is probably incomplete, feel free to contribute in order to improve or rewrite it.

Coherence : For each language there is on earth, we have computed ranked letter appearance occurrences (the best we can). So I thought that intel is worth something here. So I use those records against decoded text to check if I can detect intelligent design.

⚡ Known limitations

• Language detection is unreliable when text contains two or more languages sharing identical letters. (eg. HTML (english tags) + Turkish content (Sharing Latin characters))
• Every charset detector heavily depends on sufficient content. In common cases, do not bother run detection on very tiny content.

⚠️ About Python EOLs

If you are running:

• Python >=2.7,<3.5: Unsupported
• Python 3.5: charset-normalizer < 2.1
• Python 3.6: charset-normalizer < 3.1

Upgrade your Python interpreter as soon as possible.

👤 Contributing

Contributions, issues and feature requests are very much welcome.
Feel free to check issues page if you want to contribute.

📝 License

Copyright © Ahmed TAHRI @Ousret.
This project is MIT licensed.

Characters frequencies used in this project © 2012 Denny Vrandečić

💼 For Enterprise

Professional support for charset-normalizer is available as part of the Tidelift Subscription. Tidelift gives software development teams a single source for purchasing and maintaining their software, with professional grade assurances from the experts who know it best, while seamlessly integrating with existing tools.