Skip to content
LanguageHelper

Détecteur de langue et d’écriture IA

Collez un texte pour savoir dans quelle langue il est écrit, quelles écritures il utilise et quel outil utiliser ensuite.

0

Résultat
Télécharger
Le résultat s’affichera ici.

 

100 % dans votre navigateur

À propos de cet outil

The detector combines two signals. First it analyzes Unicode scripts — Latin, Cyrillic, Greek, Arabic, Devanagari, Han, Hiragana, Katakana, Hangul and Thai — which alone identifies languages like Japanese, Korean or Thai. Then a trigram model under 1 MB scores the text against dozens of languages written in the same script, telling Spanish from Portuguese or Russian from Ukrainian. You get a confidence score, alternative candidates and one-click links to the right tool for the language, such as the Diacritic Restorer for unaccented French or the Mojibake Fixer for garbled text.

Comme tous les outils de LanguageHelper, Détecteur de langue et d’écriture fonctionne entièrement dans votre navigateur. Votre texte n’est jamais envoyé, stocké ni utilisé pour l’entraînement, ce qui le rend sûr pour les documents confidentiels, les données clients et les messages personnels. Aucun compte ni installation n’est nécessaire, et l’outil continue de fonctionner hors ligne une fois chargé.

Comment ça marche

  1. Type or paste text — results update as you type.
  2. Read the detected language, confidence and script breakdown.
  3. Check the alternative candidates for short or mixed texts.
  4. Click a suggested tool to continue with your text pre-filled.

Cas d’usage

  • Identify the language of a document, review or message you received.
  • Route multilingual content to the right translator or tool.
  • Check whether a text is Russian or Ukrainian, Spanish or Portuguese.
  • Spot mixed scripts in names, usernames or data fields.

Exemples

Cyrillic

Avant

Привіт, як справи?

Après

Ukrainian (Українська) · Cyrillic 100%
Mixed

Avant

東京でラーメンを食べました

Après

Japanese (日本語) · Han + Hiragana + Katakana

Conseils pour de meilleurs résultats

Detection gets more reliable with more text: one full sentence is usually enough, while one or two words can be ambiguous. If a text mixes languages, the script breakdown and candidate list show the mixture. Use the suggested tool buttons to continue straight away — your text is carried over to the next tool automatically.

Questions fréquentes

How accurate is it?

For a sentence or more, accuracy is very high. Very short inputs (one or two words) are inherently ambiguous; the confidence score and candidate list show when the result is uncertain.

How many languages can it recognize?

The trigram model recognizes more than 80 major languages, and script analysis covers all major writing systems. All 20 languages of this site are fully supported.

Can it tell similar languages apart?

Yes, in most cases. Russian and Ukrainian are separated by letters unique to each (ы, э versus і, ї, є, ґ), and Spanish, Portuguese and Italian by character patterns.

Does it work offline?

Yes. The detector is a small JavaScript model bundled with the page, so it works instantly with no download and no network connection.

What are the tool suggestions?

Based on what it sees — missing accents, no punctuation, ALL CAPS, garbled encoding or a non-Latin script — the detector links to the tool most likely to help, carrying your text over.