Rephrase text

Shorter, more formal or simpler version. Free, no upload. 1 GB max · Up to 1 GB · Processed locally, never sent to a server.

Rewriting covers three different requests: trimming a text that circles its own point, lifting a register that is too casual for a professional reader, or untangling administrative prose nobody can follow. This page treats the three cases separately, with three distinct passes you select in the Objective field. None of them rewrites your sentences. Each applies substitutions and deletions whose lists are written out in plain sight in the page source.

The More formal pass rests on nineteen replacement pairs, the Simpler pass on fifteen pairs running the other way, and the Shorter pass on eight opening formulas and eight intensity adverbs that it deletes, followed by a redundancy filter. The patterns use Unicode word boundaries, precisely because the ASCII boundary in JavaScript misses accented words and would let expressions such as ça or à vrai dire slip through untouched.

So there is no language model being queried, no network request, and no text being generated. The tool moves words around inside a finite list, which explains both its reliability, since it will never head in a direction you did not write down, and its ceiling, since it will not restructure a sentence. One nuance matters above all others here: those three dictionaries are French. English input only meets the language-independent parts of the machinery, listed further down. Your original stays untouched above the result, so the two versions can be compared side by side.

How to use it

  1. Paste the text to work on The input area is never modified by the treatment. The rewritten text appears in a separate block underneath, and your original version stays available for comparison.
  2. Choose the objective Shorter (removes filler), More formal (elevated register) or Simpler (everyday vocabulary). The three passes exclude one another: only one runs per click, and the first is offered by default.
  3. Press Run The result appears immediately, with no progress bar and no remote call. Running again with a different objective replaces the block on display without passing through the previous version.
  4. Check the word count The line written into the sidebar history reports the variation, for example "25 to 15 words". A wide gap in Shorter mode means a whole sentence was judged redundant and deleted.
  5. Copy, then reread The copy button sits next to the action button and takes the contents of the result block. Rereading remains necessary: several known side effects are described below.

What each of the three objectives actually does

Shorter works by subtraction. Eight opening turns of phrase disappear, the French equivalents of "in fact", "at bottom", "to be honest", "it should be noted that" and "it is important to note that", along with three others. Eight intensity adverbs go the same way, from "très" to "complètement". Double spaces are closed up and the stray space before a punctuation mark is removed. Last comes the redundancy filter: a sentence in which more than three quarters of the words of four letters or more already appear in an earlier kept sentence is deleted outright. The rule only engages from four such long words upwards, so very short sentences are never removed by it.

More formal works by substitution, across nineteen pairs of colloquial French and its standard equivalent: "ça" becomes "cela", "du coup" becomes "par conséquent", "boulot" becomes "travail", "y a" becomes "il y a", "faut" becomes "il faut", and so on down the list. Exclamation marks are converted into full stops and sentence openings get their capital letter back.

Simpler takes the opposite road with fifteen pairs: "néanmoins" becomes "pourtant", "afin de" becomes "pour", "optimiser" becomes "améliorer", "considérable" becomes "important". This pass also does something the other two ignore. It cuts sentences at semicolons, replacing each one with a full stop followed by a capital. On a dense legal paragraph that is by far the most visible change.

The side effects to watch for

They are few, but they are reproducible. In Shorter mode, deleting an opening formula leaves the sentence starting on a lowercase letter, because the capitalization repair only exists in the other two passes. A French sentence opening with "Il est important de noter que" comes back stripped of that clause, and therefore starting on the lowercase word that used to follow it.

Still in Shorter mode, the redundancy filter reasons about vocabulary, not about meaning, and this part of the machinery works on any language. "The annual report confirms the rise in enrollments." and "The annual report contradicts the rise in enrollments." share four of their five long words, which is above the three-quarter threshold: the second is deleted, even though it states the opposite of the first. That is the one failure mode English users will meet in practice.

In More formal mode, an exclamation mark converted into a full stop keeps whatever preceded it, so French text written with its customary space before the mark ends up with a stray space before the stop. English text, which sets no space there, escapes that particular artifact and simply loses its exclamation marks. The formal and simple passes also repair French elisions everywhere, including where they changed nothing, so "le onze mai" comes back as "l’onze mai".

Why the output looks so much like the input

A rule-based tool does not move a subordinate clause, does not turn a passive into an active, does not merge two sentences and looks for no synonym outside its own list. On a text that is already clean and contains none of the targeted expressions, the output is identical to the input, and that is the normal, expected behavior.

English text is in exactly that position almost all of the time, since the pairs, the opening formulas and the elision rules were written for French. Three things still happen to an English paragraph, and nothing else:

The comparison with a language model is straightforward. A model would genuinely rewrite, reorder the ideas and adapt the tone, handing you back a text to verify sentence by sentence. Here nothing is added, so nothing needs checking beyond the side effects listed above. A substitution can still shift a shade of meaning: "toutefois" replaced by "mais" lowers the register, which is not a trivial change in a contract.

When another page suits you better

If your aim is to make a text unrecognizable, this is the wrong page: the result stays recognizable word for word. To correct spelling and grammar, the proofreading page handles that with an entirely different engine. And to cut a text in half, key sentence selection is the right tool: the Shorter pass strips filler words, it does not compress content.

Frequently asked questions

Can the meaning of my text be changed?

No sentence is rebuilt, but two mechanisms deserve attention. Substitutions occasionally shift a shade of meaning, and Shorter mode can delete a whole sentence it judges redundant. Compare the two word counters before accepting the result.

Why does my text come back unchanged?

Because it contains none of the expressions listed for the objective you chose. A text that is already plain passes through Shorter mode untouched, and a text that is already formal passes through More formal mode untouched. English text comes back unchanged more often still.

Can I chain two objectives together?

Nothing stops you pasting the result back in, but the passes partly contradict each other. Shorter mode deletes "très" and "vraiment", which More formal mode produces from "super" and restores in the form "réellement".

Does the tool work offline?

Yes. The three dictionaries and all the rules are contained in the page itself. Once it has loaded, no request is issued during processing, and nothing you type leaves the tab.

Is it any use on an English or Spanish text?

Barely. The replacement pairs, the opening formulas and the elision rules are written for French. On English, only three language-independent behaviors remain: the redundancy filter in Shorter mode, the removal of exclamation marks in More formal mode, and the semicolon split in Simpler mode. Spanish gets the same three and nothing more.

Can I undo a rewrite?

There is nothing to undo, since the input area was never touched. The result block is simply replaced on the next click, so copy a version you are happy with before running the tool again.

Will rephrasing my text get it past a plagiarism checker?

No. Rewording changes the surface of a text, not where the ideas came from, and similarity checkers flag close paraphrase rather than only exact strings. What happens on this page is far smaller than a human paraphrase in any case: words are substituted from a fixed list, nothing is written that you did not write, and on English text less happens still, since those lists were built for French. Citing the source is what settles the question; rewriting is not.

Does it use AI to rewrite the text?

No language model is queried and no text is generated. Each objective is a fixed list of replacement pairs and deletions written into the page, with a consequence you can check yourself: the same text with the same objective returns exactly the same result every time, because nothing is drawn at random. A generative tool hands back a different version on each run, which is useful when you want variety and useless when you need a result you can reproduce.

Can I paste a confidential text into it?

Nothing you paste is sent anywhere: the rules are part of the page you already loaded, and the rewrite runs without a single network request. One line does get written to your browser, though — the objective used and the two word counts — kept in local storage for the last four runs and removed by the "Clear" link beside the history title. Neither your text nor the result appears in that line, and both go with the tab when you close it.

Is there a limit on how much text I can paste in?

None is set. There is no character cap, no word count and no daily limit, since the work is a series of substitutions carried out by your own browser rather than a call to a metered service — the translation page stops at 4,500 characters for exactly that reason. The ceiling is your own machine: the pass runs on the page thread, so a very long document holds the tab still for the moment it takes.

Will an AI content detector flag the result?

There is nothing generated for a detector to find: the output is your own writing with a few listed words substituted, so it carries none of the statistical signature those tools look for. That said, AI-content detectors are unreliable in both directions and regularly flag text written entirely by hand, so no page can promise you a clean verdict. The reverse holds too — running a machine-written text through here changes almost nothing and will not disguise where it came from.

Similar tools

Discover other tools

Categories