PDFs are Paper's strongest source type. The pipeline is built specifically for long, dense documents - textbook chapters, court judgements, research papers, lecture handouts. It also accepts Word (.docx), PowerPoint (.pptx), .rtf, .epub, .txt, and Markdown files, which are converted to PDF on the way in.
Upload a PDF
- Open a page.
- In the Sources panel, click + Add source → Upload a file (or drag a PDF onto the page).
- Pick the file from your device. PDFs up to 500 MB are accepted.
What happens after upload
- Text extraction runs first, page-by-page.
- OCR kicks in automatically for any pages where the extractor got no text (scanned PDFs, image-only handouts). OCR is capped at the first 20 scanned pages per document.
- Embedding builds a vector index so the chat can quote specific passages.
- Summary and chapter detection produce a structured outline you can navigate by section.
You can keep working in the page while this happens. Status is shown next to the source - extracting, embedding, then ready. Reading the PDF works immediately; chat waits for embedding to finish.
Tips for large PDFs
- Files up to 500 MB work, but bigger files take longer - expect 1–5 minutes for a 200-page textbook.
- For textbooks and court judgements, ask the chat to "summarise chapter 3" or "explain the holding in this case" rather than the whole document at once.
- Citations link back to the source PDF page - click them to verify quotes.
If a scanned page comes through garbled, you can re-trigger extraction from the source menu. OCR runs again on pages that look empty or unreadable.
Quoting and navigation
When the chat cites a passage, the citation includes the page number. Click it to jump to the PDF at that page. Section summaries also link back to the spans they came from. Select any text in the PDF to copy it, highlight it, or pull it into chat as a quote.