How to Translate a Word Document While Keeping Its Formatting
The quick way to translate a document is to copy the text into a translation website and paste the result back. The translation may be fine, but the document is not. Bold words lose their bold, links stop working, tables collapse, and the headers and footers are left in the original language. Translating the words is the easy part. Putting them back so that the document still looks like itself is the hard part. This guide explains why formatting breaks, how a formatting-aware translator keeps it, and what you still need to check yourself.
Why Formatting Breaks in Translation
A Word document stores text in runs. A run is a stretch of text that shares the same formatting, so a sentence with one bold word and one link is split into several runs: the plain text before the bold word, the bold word itself, more plain text, the link, and so on.
Translation changes word order. In English you write "the red car"; in French it is "la voiture rouge", with the color after the noun. If "red" was bold, the bold has to move with it, to a different position in the sentence. A copy-and-paste translation throws the runs away entirely, so the formatting is simply gone. Translating run by run is worse: each fragment is translated without the rest of its sentence, so grammar and meaning fall apart.
Paragraphs also contain things that are not text at all: tab stops, line breaks, images sitting in the line, footnote numbers, page number fields. All of them have to end up in the right place in a sentence whose words have moved.
How a Formatting-Aware Translator Keeps the Layout
The approach that works is to translate whole paragraphs while marking where each piece of formatting starts and ends. Our Translate Document tool does it like this:
- Each paragraph is sent as one unit, so the sentence is translated with its grammar intact.
- Formatting is marked with tags around the words it applies to: a bold phrase, a link, a change of font or color. The translation engine places each tag around the translated words, wherever they end up, and the document's runs are rebuilt from where the tags landed.
- Objects become placeholders. Tabs, line breaks, images, footnote references and fields are carried through as markers and put back exactly where the translation leaves them, so none of them can be lost.
- Some text is protected. Web addresses, email addresses, numbers, amounts, codes and dates written in figures are kept exactly as written. A date written out in words, such as "5 March 2026", is translated whole, so it reads naturally in the new language.
- Context travels with the text. Short pieces such as table headings are translated with the surrounding text alongside them, so a word like "Bank" or "Close" is read in the sense the document uses.
What Still Changes, and Why
Even with every run in the right place, a translated document is not identical to the original:
- Length. Translations often run longer than the source. German and French, for example, usually take more space than English. Text boxes overflow, table cells wrap onto extra lines, and page breaks move.
- Fonts. A font made for Latin letters may have no characters for Arabic, Chinese or Thai. Word then substitutes another font, which can change spacing and line breaks.
- Direction and punctuation. Right-to-left languages such as Arabic and Hebrew, and languages with their own punctuation rules, deserve a careful look at alignment and punctuation.
- What is not translated. Text that is part of an image, the text in charts, and comments are not translated. A scanned PDF has no text at all, only pictures of it, so Translate Document turns scans away.
Translating With Translate Document
Translate Document accepts DOCX, DOC and PDF files up to 64 MB, with up to 100,000 characters of text, and translates into more than 100 languages. You see how much text the document holds before anything is translated.
- Choose the format you want back, a Word document or a PDF, right after uploading. The choice is locked for that upload; to get the other format, upload the file again.
- Pick the languages, and for languages that distinguish them, a formal or informal tone.
- Review side by side. Each segment appears beside its original, and the translated pages are shown as they will look. Anything that no longer fits, such as text that runs longer than the original, is cut off by its box or runs into something else, is pointed out for you to fix.
- Edit as you go. Click a line on a page to edit it, drag text to move it, use find and replace, or choose Apply original to keep a name or term exactly as it was in the source.
- Download the translated document. A bilingual Word document and a plain text version are also available.
Before you upload, know where the text goes. The file is processed on our servers and deleted automatically about five minutes after you close the page. A PDF or .doc file is also sent to a document conversion provider to prepare it for translation, and that provider states that it deletes it within three hours. The text of the document, not the file, is sent to a machine translation provider, which may keep the text and use it for its own purposes, such as improving its services. If a document is confidential, or you are not allowed to share it with a third party, do not translate it this way. Your edits are not saved either, so download the result before you close the page.
Preparing a Document for a Cleaner Result
A few habits in the original make a big difference to the translation:
- Use Word's styles for headings and body text, and paragraph spacing rather than empty lines, so that text can grow without pushing everything out of place.
- Use tables for columns, not rows of spaces or tabs, which drift apart as words change length.
- Leave room in text boxes, labels and narrow table columns for longer text.
- Keep text as text. Captions or labels baked into an image will stay in the original language.
- Finish the editing first. Accept or reject tracked changes so that there is one version of each sentence to translate.
If all you have is a scan, convert it with PDF to Word first, check the recognized text, and translate the Word file.
Machine Translation Is a Draft
Machine translation is fast and often very good, but it can mistranslate names, numbers, technical terms and negations, and it can get the tone wrong. It is not a certified or sworn translation. For legal, medical, financial or official use, have a qualified translator review it. For everything else, read the translation through before you send it, check every number and name, and if you can, ask a native speaker to read the parts that matter most.
Conclusion
Formatting survives translation when the translator works on whole paragraphs and knows where every bold word, link and image belongs. Expect text to grow, give it room, and check fonts, tables and anything inside images. Then review the translation itself, because a perfectly formatted document with a wrong number in it is still wrong.
Have a document to translate? Try Translate Document.
← Back to the blog