PDF to Word

A PDF page has no concept of "paragraphs" or "headings" the way a Word document does — internally it is a set of positioned text runs, each one saying, in effect, "draw these characters at this exact x,y coordinate, in this font." There is no outline, no document structure, not even a reliable notion of where one line ends and the next begins; a PDF viewer reconstructs the visual appearance of a page by painting every text run at its stated position, and a person reading the result perceives lines and paragraphs only because the positions happen to line up that way. Converting a PDF to an editable document means working backwards from that: grouping text runs that share a similar vertical position into lines, then noticing where the vertical gap between one line and the next is unusually large — a signal that a paragraph just ended — and treating that as a paragraph break. This is fundamentally a reconstruction from position data, not a reading of some hidden document structure, and it is exactly why the original formatting cannot be recovered: there was no structured formatting information in the file to recover in the first place, only where ink was placed.

It is easy to assume "reading a PDF" and "reading a scanned page" are the same task with the same tool, but they solve genuinely different problems. Text extraction, what this tool does, reads text that a PDF already stores as text — character codes and positions, the same data a PDF viewer uses to let you select and copy a sentence. OCR (optical character recognition) solves a much harder problem: looking at a page that is only a picture, with no text data anywhere in the file, and using image recognition to guess what letters and words the picture depicts. That is a fundamentally heavier undertaking — it involves image processing, character recognition models, and inherent uncertainty about what the "right" answer even is, none of which apply when the text is already sitting there as text. Bundling a full OCR engine into a tool whose whole appeal is running instantly in your browser would trade away the speed and simplicity that make this tool useful for the minority of PDFs that happen to be scans. A scanned, image-only PDF is a real and common case, and this tool says so plainly in its results rather than silently producing an empty document — but solving it properly means reaching for a dedicated OCR tool, not stretching this one past what it honestly does.

A more ambitious converter could try to recreate the original page's layout — placing text in boxes at approximately the right coordinates, guessing at columns, attempting to reproduce fonts. In practice this produces a document that looks superficially similar to the original but is nearly impossible to actually edit: text sitting in disconnected floating boxes rather than flowing paragraphs, reflowing badly the moment anything is added or removed, and fighting the user at every edit. Reflowing the extracted text into ordinary paragraphs instead — the approach this tool takes — deliberately gives up on visual resemblance in exchange for a document that behaves the way a Word document is supposed to: type at the end of a paragraph and it wraps normally, delete a sentence and the rest of the page reflows the way word processors have always worked. For the common goal of getting a PDF's words into an editable document — to quote, revise, repurpose or translate — that trade is the right one, and it is why this tool does not try to be a layout-preservation tool wearing a text-editor's clothing.