Free with an account · Your file stays on your device
Upload a research paper, book, or report — get a TL;DR, chapter map, key points, and quotable highlights with page numbers. We extract the text in your browser; the PDF itself never uploads.
Works with: research papers · books · reports · whitepapers · contracts · manuals
No PDF handy? Try a sample
Up to 25 MB and ~300 pages. Longer docs are summarized with the most representative portions (with a notice). Your PDF stays in your browser — pdf.js extracts the text locally, we only send the extracted text to the model.
A TL;DR, chapter-by-chapter map with page ranges, 4-8 anchor quotes with page numbers you can cite, and key points. Convert to a mind map or flashcards with one click, or paste the extracted text into the text summarizer for tone variants.
Image-only / scanned PDFs with no text layer — OCR them first, save as PDF, then come back. Password-protected PDFs — unlock before upload. PDFs with malformed text layers (bad encoding) — paste the plain text instead.
Free with an account. Sign up for 30 runs/month (each PDF = 4 credits). Upgrade to Pro for 1,500/month.
Input: 47-page research paper on transformer architectures · output shown trimmed for space.
TL;DR
The paper introduces a scaled attention mechanism that replaces recurrence with parallel token comparison, cutting training time on WMT-EN-DE from ~3 weeks to ~3.5 days while improving BLEU by 2 points. Core claim: attention alone, without RNNs or convolutions, is sufficient for state-of-the-art sequence modeling.
Chapter map
Anchor quotes
p. 4"Self-attention, sometimes called intra-attention, is an attention mechanism relating different positions of a single sequence in order to compute a representation of the sequence."
p. 10"Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results by more than 2 BLEU."
+ 3 more quotes · convert this into flashcards or a mind map in one click
No. We extract the text in your browser using pdf.js — the raw file never leaves your device. Only the extracted text is sent to our AI to produce the summary.
Text-based PDFs: research papers, books, reports, whitepapers, contracts, manuals, articles. We detect the document type automatically and adapt the summary shape.
Only if they have a text layer. Pure scans are images and need OCR first — try a free OCR tool, save the result as PDF, then come back.
Up to 25MB and about 300 pages. Longer documents are summarized with the most representative portions — we add a notice when content is trimmed.
Yes, we pass page markers to the model so quotes and chapters reference the correct pages in your original PDF.
First few summaries per day are free without signup. Free accounts get 30 credits/month. Each PDF costs 4 credits.
The tool is English-first. Summaries of non-English documents work but come out in English by default. Japanese, Chinese, Korean, Spanish, and Russian output are rolling out.