Summarise a PDF, Word or text document
Summarise a PDF, Word or text file on your device: key points, action items, tables to Excel, details found and passage search with page numbers.
Loading tool…
- The text is read on this device (pdf.js for PDF, mammoth for Word); scanned PDFs can be read with OCR.
- Summaries pick existing sentences; they do not rewrite or check facts.
- Everything runs on this device; nothing is uploaded.
Drop a PDF, a Word document (.docx) or a text file and get a quick overview without uploading it: a short summary, the key points, action items and decisions with the page they are on, the tables in the document (downloadable as CSV or Excel), the details it contains such as e-mail addresses, phone numbers, dates, amounts and links, and a search box that answers questions with passages from the document. Text is read in your browser: pdf.js for PDFs and mammoth for Word files. Scanned PDFs, where the pages are pictures, can be read with OCR (Tesseract, English and Arabic) on request.
Be clear about what this is. The summary is a basic on-device summary: it ranks the sentences of your document by how central they are (the TextRank method, on word statistics) and shows the top ones in their original order. Action items are sentences with phrases such as "must", "we need to", "please" or "by Friday" (and "يجب", "سوف", "المطلوب" in Arabic); decisions are sentences with "decided", "agreed" or "تم الاتفاق". Nothing is rewritten, so nothing is invented, but it cannot paraphrase, reason about the content or check facts. Answers to questions are passages found in your document, ranked by the words they share with your question (BM25), not generated text.
Everything is editable: fix the summary, delete a key point, add an action item, then download the report as Word, PDF (with Arabic laid out right to left) or plain text. Tables are rebuilt from text positions, so simple tables come out well, while merged cells, rotated text and tables that are pictures do not. The details found are shown on the page only; they are not stored or sent anywhere. Files up to 100 MB work; very long documents take a few seconds to analyse.
How to use it
- Drop a PDF, Word (.docx), TXT or Markdown file.
- If the PDF turns out to be scanned, click "Read with OCR" to read the pages on your device.
- Read the summary, key points, action items and decisions; each shows the page it came from, and every line can be edited.
- Download tables as CSV or Excel, copy the details found, and ask questions to find the passages that answer them.
- Download the report as Word, PDF or text.
Frequently asked questions
Is my document uploaded?
No. The file is read and analysed in your browser tab. OCR, when you ask for it, downloads the Tesseract engine and the English and Arabic language data (about 8.5 MB) from this site; the document itself never leaves your device.
Is this an AI summary?
It is a basic on-device summary: statistics pick the most central sentences and simple rules find action items and decisions. It does not use a large language model, so it never makes things up, but it also cannot rewrite, shorten sentences or understand the meaning. Treat it as a reading aid and check the original for anything important.
How does "ask a question" work?
Your question is compared with every passage of the document using BM25, the ranking method behind many search engines. The best-matching passages are shown with their page number and the matching words highlighted. They are quotes from your document, not generated answers, so use words that appear in the document.
Can it read Arabic documents?
Yes. Sentence splitting, the summary, action items and decisions work in Arabic, including common dialect phrases such as "لازم" and "هنعمل". Some PDFs store Arabic text in visual order or with split words, which can make the extracted text harder to read; OCR can help with scanned Arabic pages.
What happens to e-mails and phone numbers it finds?
They are found with patterns and shown on the page so you can copy them. They are not saved, logged or sent anywhere, and they disappear when you close the page.
Why are some tables missing or split?
PDFs do not store tables, only text positions. The tool finds runs of lines split into columns and rebuilds the grid, which works for simple tables. Merged cells, rotated text and tables drawn as images are not recovered. Tables in Word files are read directly and come out complete.
Related tools
- Extract text from a PDFExtract the text of a PDF in reading order and download it as a .txt file or copy it.
- Convert PDF to ExcelTurn tables in a text-based PDF into a real .xlsx workbook: numbers as numbers, bold header rows, sized columns.
- OCR a scanned PDFMake scanned PDFs searchable: Tesseract reads English and Arabic text and adds an invisible text layer.
- Convert PDF to WordTurn a text-based PDF into an editable Word document with paragraphs and headings rebuilt.
- AI text detector: check writing for AI signalsEstimate AI-writing signals in English or Arabic text on your device: stock phrases, sentence rhythm, personal voice and GPT-2 perplexity, sentence by sentence. Not proof of authorship.
- Create an SOP from a screen recordingTurn a screen recording into a step-by-step guide with screenshots, then download it as PDF or Word. Runs on your device; nothing is uploaded.