PDF to Word
An editable Word document from a PDF, scanned on a copier or exported from a program. The model reads the text a digital file already carries and recognises the pages that have none, so a photocopy and an export both convert. What comes back is built to be edited: real headings, paragraphs, lists and tables, with the pictures placed where they appeared, one Word page per PDF page. Up to fifty pages a run, priced before it runs.
Console
pdf-to-wordNothing is charged for uploading, and nothing runs until you press Run. Sign in to load a file and get an estimate first.
Drop your PDF in, or choose it from disk. Uploading reads nothing but the file's size and its page count, which is what the estimate is built from — no page is opened, drawn or sent anywhere until you run.
Name the pages, or leave the field alone for all of them. The pages you name are the pages the model is asked to read, and the estimate follows the count, so the number you approve is the number you pay. One run takes up to fifty pages; a longer document converts in parts.
Run and download. The pages come back as one .docx that opens in Word, LibreOffice or Google Docs, the PDF you uploaded is left exactly as it was, and a run that fails costs you nothing.
Questions
Will it work on a scan?
Yes. A page with no text of its own is recognised from the picture of it, and a page that already carries text is read from that text directly — which is both faster and more faithful than recognising it again. You do not choose between the two and the price is the same either way. Skewed or faint scans read less reliably than clean ones, and a page that is only a picture comes back as a picture.
How close to the original will it look?
Close in structure, not identical in appearance. Headings keep their levels, paragraphs and bulleted lists are real paragraphs and lists, tables are real tables you can add a row to, and each picture is lifted out of the document and placed where it appeared, with the words around it in the order they were written. Each PDF page starts a new Word page. Exact fonts, colours and column positions are not reproduced — a document rebuilt from text boxes looks right and cannot be edited, and editing is what you convert for.
What happens to headers, footers and formulas?
Running headers, footers and page numbers are left out: Word supplies its own, and a header repeated on every page would be fifty lines to delete. A formula comes through as LaTeX text in a monospace font rather than as a Word equation, so nothing in it is lost and you can rebuild it where it matters.
Can a word come out wrong?
Yes, on a page that had to be recognised rather than read: a character in fine print or a smudged scan can come through wrong — a hyphen for a dash, a lost space after a comma. In prose that is a typo you will see and fix. It is also why this is the wrong tool for a table of figures you intend to total: a misread digit does not look wrong. A page that already carried its own text is not guessed at and does not have this problem.
What does it cost?
A flat amount per run plus a fixed amount for every page you name, and a little for the pictures the document carries. The per-page part is what pays for the model's reading, which is charged the same for a scan and a digital file, so ten pages cost about ten times one. The estimate is shown before you commit, it does not change once the job finishes, and a run that fails is refunded in full.
Is there a limit on pages?
Fifty in one run. A selection over that is refused before anything is charged, with the range field as the remedy: name the first fifty, then the rest in a second run. The two documents stitch together in Word with a copy and paste.
What happens to my files?
Inputs are deleted twenty-four hours after upload and results after thirty days. Deletion means the object is removed from storage, not hidden from your history. The PDF you sent is never altered — every tool here writes a new file and leaves the one you uploaded exactly as it was.