PDF to text

Why extracted text sometimes looks scrambled

A PDF does not store lines or paragraphs. It stores glyphs at coordinates, and the order they appear in the file is whatever the program that produced it happened to write — frequently not reading order at all. Simply concatenating those fragments interleaves columns, runs words together and loses every paragraph.

Rebuild lines reconstructs the text from the geometry instead: fragments are grouped into lines by their vertical position, ordered left to right, and word spaces are inferred from the gaps between them — because many PDFs omit the space character entirely and simply position the next fragment further along. Paragraph breaks come from comparing each gap against the document's own typical line spacing.

As stored skips all of that and returns the raw order. It is worth trying when a complicated layout — a table, or several columns — confuses the rebuilt version.

Nothing came out. Why?

Almost certainly because the PDF is a scan. A scanned page is a photograph: the words are visible to you, but the file contains an image, not characters, so there is genuinely nothing to copy. Recognising them needs optical character recognition, which is a different job and one this tool does not attempt — it tells you plainly rather than handing back an empty file.

What about tables?

Table cells come out as lines of text in reading order. The columns are not preserved, because a PDF gives no indication of which fragments belong to the same cell — only where each one sits. Anything claiming to extract tables from an arbitrary PDF is guessing, and it is more useful to know that than to receive a plausible-looking wrong answer.

Is anything uploaded?

No. The document is read inside your browser and the text never leaves your device. Nothing is stored, and closing the tab discards it.