A language detector that shows its working
Paste any text and this language detector tells you what language it is in, how confident it is, and which writing system (script) it uses. It also splits a mixed text into sentences or lines and identifies each one, so an email that switches between English, French and Hindi is labelled part by part.
Two engines are available. In desktop Chrome, the browser’s built-in Language Detector API runs a model on your device and returns its own confidence score for each candidate. Everywhere else, the page uses franc-min, an open-source statistical detector (MIT licence) that compares the text’s three-letter sequences with profiles of 82 languages built from the Universal Declaration of Human Rights.
Indian users get two extras written for this page: a check for romanised Hindi (“Hinglish” such as “kya haal hai”), which a trigram detector tends to label as Iloko, Filipino or Hausa, and function-word checks that separate Hindi, Marathi and Nepali, which share the Devanagari script. A batch mode labels thousands of lines or a CSV column at once.
How to detect the language of a text
- Paste or type your text into Text; the result updates as you type.
- Read the top language, its code (BCP 47, such as
hiorpt) and its confidence, then the other candidates below it. - Check Writing systems to see which scripts the text uses, and Language of each part for mixed texts — switch between sentences and lines.
- Choose a Detector if you want a specific engine; Chrome’s may ask to download its small model once.
- For many texts, switch to Batch, paste one text per line or open a CSV and pick the column, press Detect all, then Download CSV.
Language Detector features
Confidence for every candidate
Up to five likely languages with a percentage, so a close call between Spanish and Portuguese is visible rather than hidden.
Script analysis
Letters are counted per Unicode script — Latin, Devanagari, Bengali, Tamil, Arabic, Cyrillic, Han, Hangul and 20 more.
Mixed-language texts
Each sentence or line gets its own language, script and confidence.
Romanised Hindi (Hinglish)
A word list of common Hindi words written in Latin letters spots texts such as “mujhe lagta hai” and says whether English is mixed in.
Batch and CSV
Up to 2,000 rows, a guessed text column, a language summary, and a CSV export with spreadsheet-formula protection.
How the detection works
A statistical detector such as franc counts every three-character sequence (trigram) in your text — “ the”, “ion”, “की ” — and compares their ranking with the 300 most frequent trigrams of each language. The language with the smallest difference wins, and the gap to the runner-up becomes the confidence shown here. Single-script languages such as Tamil or Korean are recognised from the script alone.
This page adds rules franc does not have. In Cyrillic, letters that a language never uses rule it out: ы and э do not exist in Serbian or Ukrainian, so a sentence containing them cannot be either. In Devanagari, words such as है and में point to Hindi, आहे and आणि to Marathi, and छ and हुन्छ to Nepali.
Scripts and their Unicode ranges
| Script | Main Unicode block | Languages (examples) |
|---|---|---|
| Latin | U+0000–U+024F | English, Spanish, Vietnamese, romanised Hindi |
| Devanagari | U+0900–U+097F | Hindi, Marathi, Nepali, Sanskrit |
| Bengali–Assamese | U+0980–U+09FF | Bengali, Assamese |
| Tamil | U+0B80–U+0BFF | Tamil |
| Arabic | U+0600–U+06FF | Arabic, Urdu, Persian |
| Cyrillic | U+0400–U+04FF | Russian, Ukrainian, Bulgarian, Serbian |
| Han | U+4E00–U+9FFF | Chinese; Japanese kanji |
When to trust the result
- Give it a sentence or more. Two or three words are often ambiguous (“no problem” is valid English and Spanish-looking), so short texts get lower confidence.
- Closely related languages — Spanish and Portuguese, Hindi and Maithili, Malay and Indonesian — are the usual mix-ups; check the second candidate.
- Names, code and numbers carry little signal; remove them from very short texts.
- To translate the text next, use the Translation Tool; to inspect unusual characters, try Text to Unicode.
Language Detector FAQ
How do I find out what language a text is in?
Paste it into the language detector. The top result shows the language, its code and a confidence percentage; the table below lists other possible languages.
Can it detect Hinglish or Hindi written in English letters?
Yes. A list of common romanised Hindi words (hai, nahi, kya, mujhe …) flags texts written that way and reports whether English words are mixed in.
Which languages can it recognise?
The statistical detector knows 82 languages, including Hindi, Bengali, Marathi, Tamil, Telugu, Gujarati, Kannada, Malayalam, Punjabi and Urdu. Chrome’s built-in detector covers a similar range.
Is my text uploaded?
No. Detection runs in your browser. Only the detector itself is downloaded — the franc-min library from jsDelivr, or Chrome’s own model.
Why is a short phrase detected wrongly?
A few words do not contain enough letter patterns to tell similar languages apart. Add a full sentence, or look at the second and third candidates.
Can I detect languages in a spreadsheet?
Yes — switch to Batch, open a CSV, choose the column with the text and press Detect all. The download adds language, code, confidence, script and a romanised-Hindi flag to each row.