Pull the existing text layer out of a PDF and keep it as plain text you can copy. Pages without text are called out. Nothing is uploaded โ copy or download when it finishes.
Drop a PDF here, or click to choose
One file, up to 50 MB. It is read in your browser.
Most PDFs already carry a text layer: the text is stored as characters, not as pictures. Files like that need no recognition at all โ the characters just have to be read back in their original lines. That is fast, exact, and needs no model. That is what this page does.
A real scan is different: the page is one big image and the file contains no characters at all. This tool does not pretend otherwise โ it lists those page numbers and says they need OCR. An exact "cannot" beats a vague "sort of extracted".
A PDF has no notion of a line, only runs of text placed at coordinates. Concatenating them gives one long string with no breaks. Two signals rebuild it: the line breaks the producer itself marked (many do), and the geometry โ a baseline drop of half a character height means a new line, a horizontal gap of a third of a character width means a space.
Tables come out as a run of text in reading order, with the columns no longer aligned โ that is a limit of plain text, not a failed extraction. Headings, page numbers and running headers land wherever they appeared. Text inside images does not appear at all.
If you need the layout, tables or charts, convert to Word or read the original โ plain text always loses structure. For scans, wait for OCR. To remove pages or merge files, the PDF toolkit is the right tool; those operations never need the text read out.
One file at a time, up to 50 MB. Everything runs in the browser: there is no endpoint that accepts a PDF, the text goes nowhere, and closing the tab discards it. A password-protected file is rejected at read time.