An AI humanizer that takes French seriously.

Why French breaks most humanizers

Run French text through a typical humanizer and you get one of three failures. Accents vanish, turning étudiant into etudiant somewhere mid-rewrite. Idioms get processed through an English brain, so expressions come back as word-for-word translations no francophone would say. Or fragments of the output simply arrive in English, because the underlying model defaulted to its dominant language under pressure. Any of these makes the output unusable; together they explain why francophone users have learned to distrust this entire tool category.

The cause is structural: those tools were built and tested on English, with other languages as an afterthought. This one was built for a bilingual country. The engine carries an explicit language preservation rule, French in means French out, and it was tested on Canadian French specifically, not just Parisian textbook French. Quebec French expressions are treated as correct French, because they are.

Two ways to use it

If you found this page, you probably work in English but write some French: reports for a bilingual employer, correspondence with francophone clients, coursework in immersion. Use the tool right here on this page; paste French text and it comes back as natural French, no settings needed. The tone options work in both languages, and the detector handles French as well, so the full detect, humanize, re-check loop works for your French drafts too.

If you would rather work in French entirely, the site has a full French side, not a translated footnote: humaniser un texte is the French homepage, humaniseur IA covers the free tier, and détecteur IA is the French detector. Same engine, same privacy, French interface, French error messages, French FAQs.

The bilingual workflow, concretely

A typical bilingual session looks like this: an English report section humanized with Professional tone, then the French summary for the Quebec office pasted into the same panel, no settings changed, returned as clean French. The engine keys on the language of each input, so mixed workdays need no toggles. For longer French documents, the same 6,000 character sizing advice applies, split at section seams. And if a francophone colleague prefers French labels and error messages, point them to the French interface; the engine is identical, only the chrome changes.