PDF를 Word로

이 도구가 하는 일

This rebuilds a Word document from what a PDF actually contains, which is glyphs at coordinates rather than paragraphs. Headings, lists and tables are inferred from the geometry of the page and written into a real .docx. It is a reconstruction, not the original recovered: the text re-flows in Word, so line and page breaks will not always fall where the PDF put them.

Use it when you have to change words in a document nobody sent you the source for: a contract that needs amending, a template letter worth re-using, a report whose figures have moved on, or a page of text you would rather edit than retype.

작동 방식

1. Drop the PDF onto this page. One file at a time, up to 200 MB.
2. Choose the layout mode. Flowing gives paragraphs that re-wrap as you edit; preserve keeps the original line breaks and looks closer to the PDF.
3. Leave heading, table and list detection on unless the document is plain prose you want kept plain.
4. Under advanced options, give a page range to convert one chapter rather than the whole file.
5. Press Convert to Word.
6. Open the .docx and check the tables first - they are where reconstruction is hardest.

A PDF does not contain the Word document it was made from. When the PDF was written, the paragraph, the heading style and the table were thrown away, and what survived was ink at coordinates. Turning that back into structure means inferring it: a line in a larger or bolder face becomes a heading, lines sharing a hanging indent become a list, text in consistent vertical gutters becomes a table. The same analysis feeds PDF to Markdown, so there is one detector to improve rather than two that would slowly disagree.

The two modes answer different questions. Flowing is for editing: sentences become paragraphs that re-wrap when you change a word, which is what makes the file useful and also why it will not look identical to the PDF. Preserve is for looking at: each line becomes its own paragraph, so the page keeps its shape and edits badly, since typing into one line pushes nothing along to the next.

A scan is checked for before any work starts. The first five pages are sampled for a text layer, and if there is none the conversion stops and points at OCR, because a scan is a picture of words with nothing in it to convert. That is a deliberate choice over handing you an empty document and letting you conclude the tool is broken.

The .docx is assembled in your browser - the OOXML written by hand and zipped locally - so no part of a confidential contract is uploaded to be converted. The only thing that crosses the network is the allowance check, which is worth knowing when the document is the sort you would not put through a web service in the first place.