Not a language model, no. It ranks and selects sentences statistically in your browser. That is a deliberate trade: less fluent than a generated summary, and incapable of inventing a fact.

Is my file uploaded to a server?

No. It is read and converted by your browser, on your machine. Load the page, disconnect from the internet, and the tool still works — which is the simplest way to prove it to yourself.

Can the summary be wrong?

It can be incomplete — it may miss something you consider important. It cannot state something your document does not, because every sentence is copied from it.

How many sentences should I choose?

Five is a good default for a report. For a long document, more sentences give better coverage; the useful limit is where it stops being shorter than skimming.

PDF Summarizer

Condense a document to its key sentences — extractive, so every sentence is verbatim from your file and nothing can be invented.

Runs in your browser — nothing is uploaded

PDF to summarise

Drop a PDF here

Nothing is uploaded — the text is read on your device.

Your file never leaves this device. The whole operation runs in your browser — nothing is uploaded, queued on a server, or deleted later, because nothing was ever sent.

Original

—

Load a PDF

Summary

—

Adjust the length

Every sentence is verbatim from your document

This is extractive: it selects the sentences that best represent the text rather than writing new ones. That means it cannot invent a fact — which is precisely the failure people fear in a summary of a contract, a report or a medical letter. Nothing is sent to a language model, so nothing is hallucinated and nothing is uploaded.

Why use it

Built to be genuinely useful

Cannot hallucinate

Sentences are selected, never written. A summary of your contract cannot contain a term your contract does not.

No account, no API

Nothing is sent to a language model, because there isn't one.

Your file never leaves the device

The conversion happens in your browser. Nothing is uploaded, so nothing has to be trusted or deleted later.

Adjustable length

Two sentences to fifteen, in the document's own order.

How it works

Three steps

1
Drop in the PDF.
2
Choose how many sentences you want.
3
Read or copy the summary.

Extractive, and why that is the right choice here

There are two ways to summarise. Abstractive summarisation writes new sentences describing the text — fluent, and capable of stating things the document does not say. Extractive summarisation selects the most representative existing sentences.

For a document you are summarising because you have not read it — a contract, a policy, a report, a medical letter — the ability to invent is disqualifying. You cannot check a summary against a document you have not read.

Every sentence here appears verbatim in your file. The summary can be incomplete, but it cannot be wrong.

How sentences are chosen

Each sentence is scored by the frequency of the meaningful words it contains, normalised so that length alone does not decide it. Pure averaging favours very short sentences; pure summing favours very long ones, so the normalisation sits between the two.

The opening sentence gets a small preference, because documents usually state their subject first.

The selection is then returned in the document's original order rather than by score, because a summary whose sentences are shuffled reads as nonsense.

FAQ

Questions people ask

Related tools

More pdf tools

Available tools are linked; the rest are in development.

Chat With PDF PDF OCR Ask Questions From PDF Image Compressor Image Resizer Image Optimizer JPG → PNG PNG → JPG JPG → WEBP WEBP → JPG PNG → WEBP GIF → JPG

All PDF Tools

Need a tool like this for your business?

We build internal tools, dashboards and automation that fit how your team works.

See Our Services