Translate a PDF and keep the formatting
Most PDF translators lose the layout because they treat the document as a bag of text. Here is what actually breaks formatting in PDF translation, how PDFlee avoids it — and where the honest limits are.
Processed on our server for this tool — your file is deleted the moment processing finishes, success or failure.
Why translation breaks PDF formatting
A PDF is not a text document — it is a set of drawing instructions: put these glyphs at these coordinates, this image here, this line there. Translation changes the text's length (German typically +25%, French and Spanish +15–25%, English targets often shorter), and the original coordinates were chosen for the original text. Naive approaches either export the text, translate it and hand you back a plain text file — layout gone entirely — or paste translated text over the original positions and let it overflow, overlap images and burst table cells.
Formatting survives translation only if the tool re-fits the new text into the old geometry deliberately. That is a layout problem, not a translation problem, and it is where PDF translators genuinely differ.
How PDFlee preserves the layout
Each output page starts as a full copy of the original page — images, vector graphics, colours and backgrounds are the original ones, not reconstructions. The areas where source text was drawn are covered, and the translated text is re-flowed into each source block's own bounding box: same position, same approximate size. When the translation is longer than the original, the font size steps down in small increments until the text fits its box, rather than overflowing into the neighbouring content.
The translated text is drawn in an embedded Noto Sans covering Latin, Greek and Cyrillic scripts with their diacritics, so the output renders identically everywhere. One honest consequence of the copy-based approach: the original text remains present in the file underneath the covering patches — visually hidden, but still there if someone extracts text from the file. If the source text itself is confidential, that matters, and we would rather tell you than have you find out.
The honest limits
No tool preserves every layout perfectly, and claims to the contrary are marketing. The specific limits here: very text-dense pages can end up with visibly smaller type, since fitting +25% more text into the same box has to come from somewhere. Damaged or unusually constructed PDFs occasionally cannot be copied page-for-page — the fallback rebuilds the text layout on clean pages, which keeps positions but loses original imagery.
PDFlee translates the text embedded in your PDF. Scanned or photographed documents — where each page is just an image — contain no embedded text, so they cannot be translated here. Run OCR first, then translate the OCR'd file.
Right-to-left targets (Arabic, Hebrew) are not offered — proper RTL text shaping is a different engineering problem, and half-doing it produces broken documents. Japanese, Korean and Chinese are available on the premium tier, where DeepL's document engine renders those scripts itself.
Cost, in the open
Pricing is in credits and shown before anything runs: 3 credits per unit on the standard tier, 5 on premium, where a unit is up to 10 pages or 20,000 characters — whichever is larger. Credits are only deducted when the translation succeeds; failures are refunded automatically. New accounts start with 3 free credits, enough to translate a typical document end to end.
Translation is the one PDFlee tool that runs on our server: your document is uploaded over HTTPS for processing and deleted the moment the job finishes, success or failure. Every other PDFlee tool (merge, split, compress and the rest) runs entirely in your browser.
Does any PDF translator keep formatting perfectly?
No — translated text is a different length than the original, so something must give. Good tools re-fit text into the original geometry and shrink the type slightly when needed; tools that promise pixel-perfect output for every document are overpromising.
What happens to images, charts and stamps?
In PDFlee they survive as-is: each translated page is built on a copy of the original page, so everything that is not text is the original artwork.
Is the original text really gone from the translated file?
No — it is covered, not removed. The source text remains in the file underneath the patches and could be extracted. If the original text is itself confidential, treat the translated file with the same care as the original.
Which languages can keep formatting?
All 26 standard-tier target languages (Latin, Greek and Cyrillic scripts). Japanese, Korean and Chinese are premium-only, rendered by DeepL's document engine. Arabic and Hebrew are not offered.