Zellio.io

PDF to Word Converter

Turn a PDF into an editable .docx — headings, tables, charts, pictures and colour kept — without uploading it.

Runs locallyWorks offlineShare link carries settings, never your data
Send output toQR Code GeneratorPassword GeneratorFocus TimerBuild a workflow from thisOutput stays on this page until you send it.

Frequently asked questions

Will the Word file look exactly like the PDF?

No, and that is deliberate. A PDF is a fixed layout; a Word file is text that reflows when you edit it. This produces the second: correct words, headings, lists, tables and pictures, in reading order, with the fonts substituted for metric-compatible ones so line lengths stay close. The appearance that carries meaning does come across — text colour, the coloured band behind a title, a tinted table header, and whether something was centred or set to the right — because a document that loses those stops looking like itself. What it does not reproduce is the fixed geometry: exact line breaks, exact page positions, and anything drawn purely for decoration. If you need an exact visual copy, you already have one — the PDF.

Can it convert a scanned document?

Not into text, because there is none: a scan is a photograph of a page. The tool detects this and says so rather than producing an empty file. To get words out of a scan you need optical character recognition — the Image to Text tool here runs Tesseract in the browser and will read a page image into text you can paste into a document.

What happens to tables?

A table drawn with ruling lines becomes a real Word table, cell by cell, with the first row marked as a header so it repeats across pages. A table held together only by alignment — an invoice, a price list, a CV — is found only if you tick the option, because that shape is indistinguishable from a hanging indent and the false positives are worse than the misses.

Does it keep colours and the look of the page?

The colours that carry meaning, yes. Text keeps the colour it was drawn in, a filled band behind a heading becomes shading on that heading's paragraph rather than vanishing, a tinted table header stays tinted, and centred or right-aligned text stays where it was put. This matters more than it sounds: a title set in white on a dark band is invisible in a converter that keeps the words and drops the band, and a report whose headings all come back black and hard left no longer reads as the document it was.

Are charts and images kept?

Yes. Photographs and logos are placed as pictures at their original size. Charts and diagrams, which are hundreds of drawn lines rather than an image, are grouped into a single picture along with their axis labels and legend, so the labels do not end up scattered through the prose. Each picture is redrawn at the resolution the original had, and a numbered caption is kept as a caption.

Is there a size limit?

Up to three hundred pages in one go, and the practical limit is your device's memory rather than any rule here: a long document with many pictures uses more than a short one. There is no queue, no daily cap and no watermark. If a file is larger than that, split it with the PDF Toolkit and convert the parts.

Can it open a password-protected PDF?

Yes, if you know the password. Type it when asked and the file is decrypted in the tab to be read. The password is used for that and nothing else — it is not stored, and it does not travel anywhere.

Which programs open the result?

Anything that reads .docx: Word 2007 and later, Google Docs, LibreOffice Writer, Pages and most mobile editors. The file is written to the Office Open XML standard rather than to any one program's quirks.

Is my document sent anywhere?

No. Everything runs locally in your browser using standard web APIs — your text, files and inputs are never uploaded to a server, so the tool works even offline once the page has loaded.

Is it free?

Completely. There is no sign-up, no account, no watermark and no usage limit. The tool is supported by unobtrusive ads, not by selling or processing your data.

Pro tips

  • Read the preview before downloading. It shows the document as Word will show it, so a column read in the wrong order or a caption swallowed by a paragraph is obvious in seconds — far quicker than finding it later in the .docx.
  • Turn off the page breaks for something you intend to rewrite, such as a report you are turning into a briefing. Leave them on when the pagination carries meaning, like a contract with clauses referenced by page.
  • If a CV or an invoice comes back as paragraphs where you expected rows, tick the borderless-tables option and convert again.
  • For a very long document, split it first with the PDF Toolkit and convert a chapter at a time — several small files are quicker to check and much easier on a phone's memory.
  • Check the verdict line before anything else. Reliable means the text came out of the file; check means something in it is odd; a scan means there was never any text to convert.

About PDF to Word Converter

A PDF holds glyphs at coordinates, not paragraphs, so a converter has to work out what the geometry means: which lines belong to one paragraph, where a column ends, which rules make a table, and which cluster of drawing is a chart rather than a border. This one shows that reading before you download it — every heading, table and picture it found, laid out as Word will lay them out — because the failure worth catching is the silent one, a lost column or a chart that became a row of stray axis labels.

Conversion happens in three passes, and the page names each one as it goes because each fails differently. The first reads every page with Mozilla's pdf.js, the same engine Firefox uses to display PDFs, collecting each run of characters with its position, its size and the font it was set in. The second decides what that arrangement means. The third crops the pictures and writes the .docx, which is a ZIP of XML parts assembled here rather than by an office suite.

The middle pass is the one that decides whether the result is worth having. Text on a facing page arrives in whatever order the producer emitted it, so runs are sorted by baseline and rebuilt into lines; lines are joined into paragraphs using the leading, the indent and whether the line before stopped short of the margin; a word split across a line break has its hyphen removed and is closed up again. Two columns are separated before any of that happens, by finding the vertical band of white that no line crosses — build the lines first and the left column's words are welded to the right column's, and nothing later can undo it.

Formatting survives at the level that matters for editing. A line set larger than the body becomes a Heading 1, 2 or 3, so the navigation pane in Word fills in and the document can be restructured rather than retyped. Bold and italic are kept run by run, from the PostScript name of the font rather than from guesswork. The families themselves are swapped for metric-compatible ones — Arial where the PDF used Helvetica, Times New Roman where it used Times — because those substitutes have identical character widths, so a line that filled its measure still fills it instead of reflowing and pushing a page of text downwards.

Tables and lists come back as tables and lists. Where a table was drawn with ruling lines, the grid is read from the lines themselves and the words are posted into cells, which is the case worth trusting: the producer has said where the cells are. The header row is marked as one, so it repeats if the table breaks across pages in Word. A bullet or a number at the start of an item becomes Word's own list numbering rather than a character in the text, which is why the bullets stay aligned when the list is edited. Tables ruled by nothing but alignment are found too, but only when you ask, because a hanging indent looks exactly like a two-column table to a computer and a wrong table is harder to fix than a missing one.

Charts, diagrams and photographs are the part most converters quietly drop. A chart is not an object inside a PDF: it is several hundred separate paths and thirty scraps of text, which is why a naive conversion produces a paragraph reading "0 100 200 300 Q1 Q2 Q3 Q4" and no chart. Here the drawing is grouped by proximity into one picture, the labels standing inside it are taken out of the running text and kept with it, a numbered caption underneath is recognised and placed as a caption, and the whole region is redrawn as an image at a resolution taken from the file's own pixels. Photographs come out at their own density rather than at a guess.

What the tool will not do is pretend. A PDF made by scanning paper contains pictures of words and no text at all; no rearrangement of geometry can recover them, and the verdict panel says so instead of handing back an empty document. A file whose fonts lack the table that maps glyphs back to characters produces mojibake, and that is reported as a percentage rather than shipped quietly. Text colour, footnotes, equations and form fields are not carried across yet. The aim is a document whose words, structure and pictures are correct and editable — not a pixel-perfect impersonation of the original, which is what the PDF itself already is.

Common use cases

  • Editing a report that arrived as a PDF when the original document is long gone.
  • Lifting the tables out of a research paper or an annual report to work on the numbers.
  • Reusing the text of a contract or a policy as the basis for a new version.
  • Getting a CV back into an editable state after it has been through a PDF export.
  • Rebuilding a presentation handout or a newsletter whose source files were never kept.
  • Pulling the charts out of a report as pictures for a deck, with their captions intact.
How it comparesAdobe's own converter and the familiar free sites — Smallpdf, iLovePDF, ILovePDF's many imitators — run the conversion on a server, usually a copy of LibreOffice, and they are good at it: their layout fidelity on complex documents is better than anything achievable in a tab today. The trade is the upload, plus the daily task limits, the sign-in and, on some, a wait. This one converts on the machine you are sitting at, with no file size cap, no account and no queue, and it shows you its reading before you commit to it. For an ordinary report, contract, CV or paper the results are comparable; for a heavily designed magazine page, a server-side converter will still do better.