Help›Word (.docx) Compatibility

Word (.docx) Compatibility

Last updated

Word Compatibility in Revise

Our mission with Revise is build the most advanced agentic word processor in the world, and as such we treat Word files (.docx) as a first-class format. Our goal is that a document can pass through Revise on its way between Word users and nothing gets lost. This page is a current map of what survives the round trip today, what is partially supported, and what is still on the roadmap. We update it as coverage grows. (Under the hood, .docx is the Office Open XML format, standardized as ECMA-376 — a zip package of XML parts, described well in Microsoft's WordprocessingML overview. The feature links on this page point at the underlying OOXML elements for readers who want the spec-level detail.)

If you find a bug with our .docx format support, please reach out to support@revise.io and let us know!

Byte-for-byte round-trip fidelity

Most editors that open a Word file rebuild it from scratch on export — every part of the file gets rewritten, whether you touched it or not. Revise doesn't. When you import a .docx, the original file stays with the document, and on export everything you didn't change comes back as the exact bytes that arrived: same XML, same formatting Word wrote, no reformatting, and no added bookkeeping markers. Only the paragraphs you actually edited are re-written. Import a 90-page contract, fix one clause, and the other 89 pages leave Revise untouched — byte for byte.

This preservation extends past what Revise itself can display:

  • Formatting we don't model still survives editing. Word (and its plugins) attach properties Revise doesn't render — vendor text effects, revision-save identifiers, table look flags, page borders, and the w14/w15 extension properties Microsoft layered onto the standard. These ride along invisibly and are re-emitted on export, even inside paragraphs you edited.
  • Objects we can't draw are preserved, not dropped. Embedded spreadsheets and other OLE objects, SmartArt, charts, and even OOXML constructs we've never seen before appear in the editor as labeled placeholder chips — you can see them, move them, or delete them — and export back to Word exactly as they came in.
  • Editable-region permissions survive. Word's restricted-editing ranges (w:permStart / w:permEnd) round-trip with their pairing guaranteed: typing inside a region extends it, and deleting its text retires it cleanly.
  • Document properties are never rewritten. The file's title, author, and creation date come back exactly as they went in — editing a document doesn't change who created it or when.

A verification stage inspects every exported file before it leaves; in the rare case something can't be verified, Revise falls back to a fully regenerated (still valid) Word file rather than ship anything questionable. Fidelity degrades gracefully — it never corrupts.

Fully supported (round-trips both ways)

  • Text and formatting - paragraphs, headings, alignment, line spacing, indents, and page breaks; bold, italic, underline, strikethrough and double strikethrough, small caps, all caps, character spacing, highlights, text colors, font sizes and families, superscript/subscript, hidden text, and links. Documents that carry their fonts through Word's theme (most do) resolve to the right typefaces.
  • Paragraph borders and shading - per-edge borders (single, double, dashed, colored) and background fills, rendered on screen and in PDF.
  • Tab stops - left, right, center, and decimal alignment, plus dot, hyphen, and underscore leaders. There is no ruler - ask the agent to set or move stops.
  • Watermarks - DRAFT / CONFIDENTIAL-style text watermarks import from and export as Word's native format, and there's a dialog on the Layout tab to set your own.
  • Line numbers - margin line numbering (pleading paper), with restart and count-by options, on screen and in PDF.
  • Pagination controls - keep-with-next, keep-lines-together, widow/orphan control, and automatic hyphenation all paginate the way Word does.
  • Tables of contents - a TOC imports as a live block that tracks your headings, and exports with real page numbers in both .docx and PDF.
  • Clause numbering and cross-references - Word-style numbering schemes ("Section 1.", "1.01", "(i)") stay live - insert or delete a clause and everything renumbers - and REF / PAGEREF cross-references keep pointing at their targets.
  • Tracked changes (w:ins / w:del) - insertions, deletions, formatting changes (with revert-to-prior on reject), paragraph split/merge revisions, moved text, table-row revisions, and paragraph- and table-property revisions (reject reverts the prior properties). Author names, dates, and Word revision identity are preserved in both directions, and Word redlines arrive in Revise as reviewable suggestions.
  • Footnotes and endnotes - both streams, with correct per-stream numbering.
  • Headers and footers - including three-zone layouts, different first-page headers, even-page slots, and PAGE / NUMPAGES fields.
  • Comments - anchored ranges, reply threads, and resolved state are preserved through import and export, including Word's original comment IDs and ranges spanning paragraphs. Word comments arrive as live comment threads in Revise's margin - you can reply, resolve, and add new comments, and it all exports back to Word.
  • Lists - ordered, unordered, and checklists, with basic nesting.
  • Tables - including merged cells, cell backgrounds, column widths, and header rows.
  • Inline images (PNG, JPEG, GIF, BMP, SVG), equations (Word math converts to LaTeX and back), visible text in DrawingML and VML text boxes (flattened into editable paragraphs in reading order when floating geometry cannot be represented), and block-level content controls (the wrapper - tag, alias, and bindings - survives the round trip; inline controls keep their content but lose the wrapper).
  • Right-to-left text - RTL and bidirectional paragraphs (Hebrew, Arabic) import, render, and export to Word, PDF, and HTML.
  • Page setup - size, orientation, and margins.
  • Multi-section documents - per-section page size, orientation, margins, page-number restarts, and per-section headers & footers round-trip and render (a landscape exhibit stays landscape, with its own header). Continuous boundaries and 2- or 3-column sections render in Revise and export to Word, PDF, and HTML.

Partially supported

These import safely - the text always survives - but some structure or positioning is simplified:

  • Text boxes and shapes - the text inside them imports as ordinary paragraphs; the box geometry itself is dropped.
  • Floating images - the image survives but becomes inline; position and text-wrap settings are lost.
  • Named styles - paragraphs remember their style identity and the document's original style definitions (firm templates included) survive the round trip, so Word shows "Heading 1" or "Firm Body Text" - not baked-on overrides. Editing the style definitions themselves inside Revise isn't supported yet.
  • Legacy vector images (EMF / WMF) - preserved byte-for-byte through the round trip and re-exported intact; Revise renders a labeled placeholder on screen and in PDF (browsers can't draw these formats).
  • Fields and bookmarks - PAGE / NUMPAGES, TOC, and REF / PAGEREF cross-references stay live (see above); other field types freeze to their last-cached text.

Not yet supported

  • SmartArt and chart rendering - these objects survive the round trip byte-for-byte and appear as labeled placeholders in the editor (see the fidelity section above). Drawing their contents on screen, and editing the text inside them, is not supported yet.

Roadmap

Our priority order is simple: anything that could silently lose content gets fixed first — that tier is clear: the byte-preserving round trip keeps even features Revise cannot display intact through export. Next comes structure (merged cells, sections, tab stops, borders, and watermarks are done; floating-image positioning is next), then semantics (SmartArt and chart rendering, style editing, more live field types). Every release is checked by an automated round-trip suite that pushes generated torture-test documents through import and export and verifies nothing is lost. If a gap on this page matters to your workflow, tell us - real documents from real users are how we prioritize.