How to Translate a PDF: 7 Ways That Actually Work (2026)
Published 2026-07-09
The short answer: to translate a PDF for free, upload it to Google Translate's document tab — it's fast but destroys formatting. To translate a PDF and keep the layout, tables, and formulas intact, use a layout-preserving AI translator like KolmoPDF. Everything between those two options is a trade-off, and this guide walks through all seven of them.
Which method should you use?
| Method | Free? | Keeps layout? | Scanned PDFs? | Best for |
|---|---|---|---|---|
| Google Translate | Yes | No | Partially | Quick gist of a short document |
| Microsoft Word | With license | Partially | No | Simple business documents |
| Google Docs | Yes | No | Partially | Casual use, short files |
| DeepL | Limited | Partially | Pro only | European languages, business text |
| ChatGPT / Claude | Limited | No | Yes | Small excerpts, nuanced tone |
| Layout-preserving AI (KolmoPDF) | Free credits | Yes | Yes | Technical docs, papers, manuals |
| Human translation | No | Yes | Yes | Legal / certified documents |
1. Google Translate (free, fastest)
Go to translate.google.com, switch to the Documents tab, upload your PDF (up to 10 MB), pick the target language, and download the result.
The catch: the output is a plain-text reconstruction. Multi-column pages get scrambled, tables collapse, images disappear, and math becomes unreadable. For a two-page letter it's fine; for a datasheet or paper it usually isn't.
2. Microsoft Word
Word can open many PDFs directly (File → Open), converting them to editable text. Then use Review → Translate → Translate Document.
The catch: Word's PDF import is unreliable for anything with complex layout, and it can't handle scanned pages at all. Expect to spend time fixing formatting after translation.
3. Google Docs
Upload the PDF to Google Drive, open it as a Google Doc (this runs basic OCR), then Tools → Translate document.
The catch: same story — the OCR-to-Doc step flattens the layout before translation even starts. Footnotes, headers, and tables rarely survive.
4. DeepL
DeepL produces some of the most natural translations available, and its document mode accepts PDFs. Formatting preservation is better than Google's but still struggles with dense technical layouts, and PDF support with OCR requires a paid plan.
Best for: business documents between European languages where tone matters.
5. ChatGPT or Claude
Upload a PDF to ChatGPT or Claude and ask for a translation. Quality is excellent for prose — these models handle idiom and context better than classic machine translation.
The catch: output is chat text, not a document. You lose all layout, page structure, and figures; long PDFs hit context limits and get summarized rather than fully translated. Good for translating a few paragraphs precisely, bad for delivering a translated document.
6. Layout-preserving AI translation
This is the category built for the hard cases: academic papers, technical manuals, textbooks, anything with formulas, tables, and multi-column layouts.
KolmoPDF's PDF translation parses each page with a Visual Language Model first — reading it the way a human does — then translates and re-renders the document so the translated PDF keeps the original structure:
- Upload the PDF (native or scanned — OCR is built in).
- Choose from 8 target languages: English, Chinese, Japanese, Korean, French, German, Spanish, Russian.
- Download a translated document where LaTeX formulas, tables, and reading order are intact.
New accounts get free credits, so a typical paper translates free. If you need the same thing programmatically, there's a PDF translation API.
The catch, honestly: per-page pricing beyond the free tier, and stylistic prose polish is tuned for technical content rather than literary flair.
7. Human translation services
For legal contracts, immigration documents, and anything needing certification, machine translation isn't accepted. Expect roughly $0.08–0.25 per word and turnaround measured in days.
How to translate a scanned PDF
A scanned PDF is photos of pages, so text-based tools see nothing to translate. You need OCR before (or built into) translation. Fastest path: a tool that OCRs and translates in one pass — KolmoPDF treats every page as an image anyway, so scanned and native PDFs go through the same pipeline. Alternatively, run OCR in Google Drive first, accept the layout loss, and translate the result.
How to translate a PDF to English (or from any specific language)
All seven methods above are direction-agnostic, but quality varies by language pair. Machine translation into English is generally stronger than out of it; for Chinese/Japanese/Korean technical documents, layout-preserving tools have a practical edge because those documents are heavy on multi-column layouts and figures that plain-text pipelines scramble.
FAQ
Can I translate a PDF for free? Yes — Google Translate handles files up to 10 MB free, and KolmoPDF's signup credits translate a typical document at no cost. "Free" tools differ mainly in what they do to your formatting.
What's the best way to translate a large PDF (100+ pages)? Chat assistants will truncate it and browser tools time out. Use a document-native service with per-page processing, or split the file with a PDF splitter and batch the parts.
Does translating a PDF keep the formatting? Only if the tool is built for it. Text-extraction pipelines (Google, Docs, chat AI) discard layout by design; layout-preserving translators reconstruct the page.
Can AI translate PDFs with math formulas? Generic tools mangle math. Tools that recognize formulas as LaTeX during parsing — KolmoPDF, or Mathpix followed by separate translation — keep equations meaningful.