PDF & Word Summariser

How to summarise a PDF, Word document or text

1. Paste the text into Text to summarise, or click Upload file to load a .txt, .pdf or .docx document.
2. Choose a Summary length (short, medium or long) and a Format (one paragraph or bullet points). You can switch between paragraph and bullet points afterwards without summarising again.
3. Click Summarise. The first time, your browser downloads an AI summarisation model (about 275 MB), which it then caches, so later summaries start straight away.
4. Read the summary, then Copy it or Download it as a .txt file.

Everything runs on your device. The document text and the summary are never uploaded; only the one-time AI model download comes from the internet.

How the summary is written

This is a real AI summariser, not a list of the first sentences: it uses DistilBART, a transformer model trained to write new sentences that capture the main point of a passage, running fully in your browser with transformers.js. Long documents are split into sections, each section is summarised, and for multi-section documents the section summaries are combined and summarised once more so the result reads as one piece rather than a set of disconnected fragments.

Because it generates new wording rather than copying sentences, an AI summary can occasionally simplify a nuance, miss a secondary point, or very rarely state something slightly differently from the source. Treat it as a fast first read, and check the original document before relying on a number, date or quote.

Choosing a length and format

Paragraph reads as connected prose. Bullet points split the same summary into short points, which is quicker to scan; switching between the two after summarising is instant, since it only reformats the text already produced.

What works best

• English text. The summarisation model was trained on English news and reports, so it works best there; other languages may produce a weaker or partly English summary.
• Real prose. Tables, code, reference lists and very short notes do not summarise well, because there is little sentence-level content to condense.
• A few hundred words or more. Very short input (under about 40 words) is rejected, since there is nothing meaningful to shorten.

If a PDF imports with broken lines or no text, try PDF to Text first to see what text pdf.js can read from it; a scanned PDF has no text layer at all and needs OCR, for example with Image to Text, before it can be summarised.

Frequently asked questions

Is my document uploaded to summarise it?

No. The text is read in your browser, and the AI model that writes the summary also runs in your browser. Nothing about the document leaves your device; the only download is the AI model itself, the first time you use the tool.

Why does the first summary take longer than the next one?

The first run downloads the AI model, about 275 MB, which your browser caches. Later summaries reuse the cached model and start immediately, as long as you use the same browser and have not cleared its site data.

Can it summarise a scanned PDF or an image?

Not directly. A scanned PDF is a picture of each page with no text to read, so convert it with Image to Text (OCR) first and paste the result here.

How long can the document be?

There is no fixed limit: longer documents are split into sections, each summarised and then combined into one final summary, so books and long reports work, although it takes correspondingly longer and several passes through the model.

Can I get bullet points instead of a paragraph?

Yes. Choose Bullet points in the Format setting before or after summarising; switching formats afterwards is instant and does not summarise again.

Does it work in languages other than English?

The AI model is trained on English text and gives its best results there. It can still run on other languages, but the summary may be weaker, shorter than expected, or partly in English.

Updated 2026-10-18