Convert DOC to ODT

The binary Word format dates from Word 97 and OpenDocument was published in 2005, with Microsoft’s own replacement format arriving between them in 2007. Converting a .doc straight to ODT therefore skips a generation rather than stepping through it, and there is no reason to make the intermediate stop: the same engine reads the old file and writes the new one either way.

What makes the direct route sensible is that the destination is a decision about the next twenty years, not about this week. If the archive is being moved at all, it is being moved once. Landing it on the format you actually intend to keep is cheaper than landing it on the one that happens to be nearest. It also avoids a second round of verification: every migration needs somebody to open a sample of the output and confirm it is right, and doing that twice for the same archive is the kind of work that gets skipped the second time.

OpenDocument is maintained by OASIS and published as ISO/IEC 26300. Nobody has to keep a licence current to read it, no single company decides when support ends, and the specification is available to anyone who wants to write a reader from scratch in forty years. That is the whole argument, and it is a procurement argument rather than a technical one.

It is also why the requirement usually arrives as a rule rather than a preference. Submission portals for several national and regional administrations name ODT specifically, and no amount of explaining that a DOCX is also documented will change what the form accepts. If a policy is what brought you here, the policy is the reason and this is the conversion it asks for. The counter-argument, which is fair, is that a format nobody in the office can open is not preservation either. OpenDocument avoids that trap because the software that reads it is free, cross-platform and does not expire, which is the practical half of the same argument.

The awkward part of a legacy migration is usually the reading, not the writing. A binary .doc is not a format you can open with a text editor and repair by hand, and current Microsoft Office declines the very oldest variants by default. The set of programs that still read it well is small, and LibreOffice is the most capable member of it.

That is what runs here. The file is opened by the same office suite that would open it on a desktop, and written out as OpenDocument by the program for which OpenDocument is the native format. No Microsoft licence is involved anywhere in the path, which for a school or a small charity is frequently the deciding factor. It also means the conversion can be repeated later. If a better route appears, or if somebody finds a fault in a batch, the originals can be run through again — which is only true while the originals still exist and while something can still read them.

Everything that made it a document. Paragraph and character styles map onto OpenDocument styles, tables keep their structure and their merged cells, images stay anchored where they were, headers, footers and page numbering carry across, footnotes and endnotes stay attached to their references, and section breaks with their own page setup survive as sections.

Fonts are handled by substitution rather than by embedding, and the machine performing the conversion carries metric clones of the common Microsoft typefaces — Liberation for Arial, Times New Roman and Courier New, Carlito for Calibri, Caladea for Cambria. Those replacements occupy the same space as the originals, which is why page counts usually do not move. Documents from this era also frequently name fonts that were never widely distributed — a typeface bundled with a printer driver, or one bought for a single publication — and those are substituted with something approximate. It is worth noting which documents they are, because that is where a page count changes.

The differences are concentrated in features Microsoft itself later abandoned. The old drawing layer — grouped shapes, floating frames, the WordArt of that era — was reimplemented between Word generations and has no exact OpenDocument counterpart, so a document whose layout is built out of those objects can shift. Some field codes have ceased to exist entirely.

None of this touches a report, a set of minutes, a letter or a policy document, which is what most archives are made of. A rule of thumb that holds up well: if the file would already look odd when opened on a colleague’s different version of Word, it is also the file to inspect after converting. Everything else can be converted in bulk and trusted.

Binary Word documents carry a summary block that owners routinely forget about: a title that was never updated, the name of the person who last saved the file, an organisation that has since been renamed, sometimes a revision count that says more about the drafting process than anyone intended to publish.

Conversion carries the basic properties across rather than inventing or erasing them, so if this archive is going to be published rather than merely stored, open the document properties on a sample of the output and decide what should stay. It takes a minute per file and it is the sort of check that is only ever obviously necessary afterwards. For a large archive the sensible version of this is a policy rather than an inspection: decide what the properties should say, set them consistently on the converted set, and record that decision alongside the collection so that whoever inherits it knows what was changed and what was original.

A .doc can contain Visual Basic for Applications written against Word’s object model. LibreOffice has its own scripting system and its own object model, and there is no faithful translation between them, so the automation is dropped while the text and formatting arrive intact.

In a records migration this is almost always the correct outcome. An archive holds documents, not programs, and a stored record that executes code when it is opened is a liability rather than a feature. Where the automation genuinely was the document — a claim form, a calculating template — it has to be rebuilt deliberately, and that is a project rather than a conversion.

Migrations arrive as folders: a decade of committee papers, a school’s worksheets, a solicitor’s precedents. Drop the whole directory and each file converts in turn, keeping its filename with the extension replaced, coming back as a single ZIP that unpacks over a copy of the original structure.

Sample the output deliberately rather than randomly. Open the longest file, the one with the most tables, and the one whose layout somebody was proud of, because those are where a conversion fails and the plain memos are not. Retire the originals only after that, and only once the ODT copies are somewhere backed up. In most collections the honest answer is not to retire them at all. Binary Word files are small by modern standards, and keeping the source alongside the converted copy costs a few gigabytes and removes the possibility that a migration you cannot repeat turns out to have been imperfect.

This pair runs on a server, not in your browser, and the page says so rather than leaving it to be assumed. Interpreting a compound binary Word file requires a full office suite; there is no small library that does it correctly, and pretending otherwise would be the kind of claim this site does not make.

The file travels over an encrypted connection to a container running LibreOffice headless, with no outbound network access of its own. Each job gets an isolated scratch directory that is deleted the moment the job ends, whatever the outcome, and any job still running after sixty seconds is killed rather than left to hang. The free limit is 25 MB per file.

ODT is for records that must remain workable — something a person will amend, extend or reuse. If a document’s content is settled and the only remaining requirement is that it stays exactly as it is and renders the same everywhere, a PDF is the more honest container for it.

Many archives end up holding both: the OpenDocument file as the editable master and a PDF as the fixed rendition of a particular version. That costs one extra conversion and removes the argument about whether the copy in the repository is the same document as the one that was approved.