How to translate a PDF without breaking its layout
Translating a PDF means doing two things at once: rendering the text in another language, and rebuilding a document that still looks like a document. Most online PDF translators get the first part right and the second part wrong — they hand you back a wall of plain text. Here is how a layout-preserving translation actually works, what to watch out for in legal and academic documents, and which language pairs need extra care.
How layout-preserving translation works
We extract the text along with its layout coordinates and font metrics, group it into translation-coherent units (a paragraph, a table cell, a heading), translate each unit while keeping the document's structural context intact, and then re-typeset the translated text into the original layout boxes. When the translation is longer than the source — common in English-to-German or any-to-Spanish — we adjust line spacing or scale font size by 2–3% so the text still fits its box without breaking out.
Embedded images, vector charts, and diagrams are left untouched. If they contain text inside the image, that text is OCR'd, translated, and overlaid back onto the image with as close a font match as we can manage. Fonts that don't support the target language's script — for example a Latin-only font on a Russian translation — are substituted with a Cyrillic-capable equivalent.
When this tool is the right choice
Layout-preserving translation matters in three specific situations:
- Legal documents and contracts. Page references, exhibit numbers, and signature blocks need to land on the same physical pages as the source for cross-referencing. Plain-text translation breaks all that.
- Academic papers and theses. Figure callouts, equation labels, and citation numbers depend on layout. Translated layout-preserved papers can be filed alongside the originals without re-numbering.
- Onboarding and HR handbooks. Side-by-side English/local-language handbooks are often a regulatory requirement. Layout-matched translation makes them comparable page-by-page.
- Technical manuals. Diagrams with callouts must keep their callouts in the same positions. We re-overlay translated callouts on the original diagrams.
- Instruction sheets and product packaging. Exact print-area constraints mean text must fit specific boxes. We auto-fit; plain-text translators don't.
Common mistakes
These are the patterns that produce a translation you can't actually use:
- Translating before OCR on scanned PDFs. If the source has no text layer, the translator has nothing to work with. OCR first, then translate — or use the combined option that does both in sequence.
- Translating into a language with the wrong script. Going to Arabic or Hebrew flips the reading direction. Tables and bulleted lists need to be mirrored, which we do automatically — but check the result.
- Translating jurisdiction-specific legal terms. 'Consideration' in a contract has no Spanish equivalent because the concept doesn't exist in civil law. Have a bilingual lawyer review legal translations even when the layout is perfect.
- Mixing source languages. If the document mixes English with quoted German, we autodetect at paragraph level. Quoted technical terms that should stay in English need to be flagged in the Options panel as 'do not translate'.
- Trusting the translation for high-stakes documents. Machine translation is excellent for understanding and for first drafts. It is not a substitute for a sworn translator on documents that need certification.
Tips for a clean translation
Get a noticeably better result with these adjustments:
- Choose the right tone in the Options panel. Formal vs. neutral vs. conversational changes how 'you' is translated and how nouns are gendered in romance languages. Match the tone of your source.
- Provide a glossary for repeated terms. Brand names, product codes, and proper nouns can be locked so they aren't translated. Upload a CSV with source/target pairs.
- Translate one direction at a time for parallel docs. If you need both English and Japanese versions, start from the same canonical source and translate to each. Don't translate Japanese-from-English back to English to check — round-trip degrades quality.
- Spot-check the layout on the first three pages. If the first three are clean, the rest will be too. Most layout issues are caught early.
- Bundle as a multilingual handbook macro. If you need 8 languages, run the Multilingual Handbook macro once instead of 8 separate translations — same source, parallel outputs, consistent terminology.
Limits and supported languages
We support 90+ languages including Spanish, French, German, Italian, Portuguese (Brazil and Portugal), Dutch, Swedish, Polish, Russian, Ukrainian, Greek, Turkish, Arabic (MSA), Hebrew, Hindi, Bengali, Tamil, Vietnamese, Thai, Indonesian, Malay, Japanese, Korean, Simplified and Traditional Chinese, and many more. The free preview translates the first 5 pages so you can sanity-check quality before paying. Pricing is 1 credit per page per target language; a 20-page doc into 3 languages is 60 credits.