PDF to Text

PDF to Text extracts the text from a PDF and gives it to you as plain text you can copy or download as a .txt file. It is for quoting from a report, pasting a contract clause into an email, feeding a document into a spreadsheet or another tool, or simply getting the words out of a PDF that is awkward to select. The PDF is read in your browser and never uploaded.

It works on PDFs that contain real text — anything exported from Word, Google Docs, a web page or most accounting software. Inside such a file, text is stored as pieces of characters, each placed at a position on the page; there are no lines or paragraphs as such. The tool reads those pieces with pdf.js, Mozilla's PDF library, and rebuilds the reading order from their positions: pieces whose baselines are within half a font size of each other form one line, read left to right; a horizontal gap wider than about 0.15 of the font size becomes a space; and a vertical step more than 1.3 times the page's usual line spacing becomes a blank line, which marks a paragraph. Pages are separated by a blank line, and "Mark page breaks" adds a "--- Page N ---" line before each page. Pages stored sideways are read the way you see them.

A scanned PDF is different: each page is a photograph of paper, so there is no text to extract — only pixels. The tool detects pages with no text at all and says so, naming them. For those, save the pages as images with PDF to JPG and run them through Image to Text, which uses optical character recognition (OCR). This tool does no OCR, so it never guesses at words.

Layout is approximated, not reproduced. Text in two side-by-side columns, or in table cells on the same row, is read straight across, so the columns are interleaved line by line. Hyphenated words at line ends are kept as they are. Hebrew and Arabic are stored in a PDF in visual order, so each line is put back into reading order with a simplified version of the Unicode bidirectional algorithm: right-to-left words read right to left, while numbers and English words inside them stay left to right. Complex mixed-direction text (nested quotes, brackets) may still come out in the wrong order. Some PDFs store characters in an order or encoding that does not map back to readable text (certain custom fonts do this); the output then contains the wrong characters, and only OCR can help. Headers, footers and page numbers are included, because to the PDF they are text like any other.

You can choose all pages or a range such as 1-3, 5. The result shows the word and character counts, a Copy button and a Download .txt button; the file is saved as UTF-8 and named after the PDF. PDFs up to 200 MB are accepted; ones that need a password to open are refused with a clear message. Nothing is uploaded — you can confirm in your browser's developer tools (Network tab) that no request carries your file; the site loads its own scripts and analytics, never your documents.

1. The file size is checked before reading (200 MB limit) and its first bytes for the PDF signature; pdf.js then opens it. 2. For each chosen page, pdf.js getTextContent returns positioned text pieces; each is converted to the page as displayed (rotation applied, y measured from the top), with its font size from its transform. 3. Pieces are sorted top to bottom, and grouped into one line when their baselines differ by no more than half the smaller font size; each line is sorted left to right. 4. Within a line, a space is inserted where the gap between two pieces exceeds 0.15 × the font size (unless one already has a space). 5. Between lines, a step greater than 1.3 × the page's lower-quartile line spacing (or 1.8 × the font size when the page has fewer than four lines) is a paragraph break. 6. A line containing Hebrew or Arabic letters is converted from visual to reading order (right-to-left runs reversed; in a mostly right-to-left line, the run order reversed too). 7. Pages are joined with a blank line (optionally with "--- Page N ---" markers); pages with no text are listed separately.