Online summarisers almost all require uploading the document to their servers, then forwarding it to a third-party AI. For internal minutes, a contract or a medical file, that amounts to publishing the document to save three minutes.
How it works
Drop your file on the page, or click to pick it from your folders.
Start the job. It runs inside your browser: nothing is sent to a server.
Download the result. No account, no watermark, no queue.
When this really helps
Size up a report
Know within a minute whether the forty pages deserve a full read.
Prepare for a meeting
Pull the salient points out of the minutes before going back to them.
Sort through a file
Review twenty documents without opening every one.
What the conversion keeps, and what it loses
Working inside a browser comes with limits a hosted service does not have. Here they are in plain words, so you know what to expect before you start.
What is kept
- Sentences exactly as the document writes them
- The original reading order
- Figures, dates and proper nouns, untouched
- The split into paragraphs
What is not carried over
- Tables, images and captions
- Layout, headers and footers
- Passages judged secondary, by design
- Footnotes, swallowed into the body text
Worth knowing about the format
The default method is extractive: every sentence is scored on the words it shares with the rest of the document, and the best-scoring ones are returned in their original order. That is deliberate — on a document that commits you, a plausible but inaccurate rewording is worse than a faithful extract. The written mode relies on distilbart-cnn, a model trained on English only.
Not the right tool?
- Scanned PDF to text — Pour récupérer le texte intégral d'un scan plutôt que son résumé.
- Translate a PDF — Pour lire un document étranger avant de le résumer.
- PDF → Word — Pour reprendre le document entier plutôt que son résumé.
Why it's free
Online converters receive your file on their servers, process it, then promise to delete it. Every document costs them computing time: that is why they charge, and why they cap file sizes.
Here, the work happens on your own computer. The page contains no server code: your document cannot be received, because it is never sent. So there is nothing to bill, nothing to cap, and nothing to delete afterwards. One exception here: the language model, fetched once and then kept in your browser cache. It is the only network exchange on this page, and it runs that way round — your document never goes up.
Frequently asked questions
Is my document sent to an AI?
No. The default method is a statistical computation that runs in the tab, with no network. The written mode downloads a model onto your machine and runs it there: the model weights come down, your document never goes up.
Does the summary invent sentences?
Not with the default method: every sentence of the summary appears word for word in the document. That is why it is offered first.
Does it work on a scanned PDF?
Yes. If the document holds no readable text, character recognition starts on its own before the summary, on your own machine. Allow a few extra seconds per page.
Which length should I pick?
“Medium” suits almost everything. “Short” helps decide whether a document deserves reading; “Detailed” keeps a usable record without rereading the original.
Other tools
Edit the text in a PDF — Fix a line that is already written, even on a scanned document. This is the feature other services charge for, and the natural starting point when you need to retouch before converting.