File formats · PDF
Translate a PDF without losing formatting
Upload the PDF you have and get a PDF back, with the translated text written into the original page layout: the same columns, tables, headings, lists and links, the images where they were. Where the translation runs longer, the type scales to fit, so you translate a PDF without losing formatting. Machine translation runs on models we host, every segment is checked, and a person decides what is final. Files with selectable text, on every plan, including the free one.
How it works, in three steps
- Upload the PDF. The pipeline reads the page geometry first: columns and reading order, tables, lists, headings, running headers and footers, and the document tags when the PDF carries them.
- Translate in the CAT editor. Each paragraph and table cell is a segment, with translation memory, termbase, machine translation and live checks on the row. A positioned page preview shows the PDF as it is while you work (how the preview works).
- Download the PDF. The translation is written back into the original frames. Bold, italic, links and the table of contents come with it, and the type scales down wherever the translation needs more room than the source.
A PDF comes back as a PDF, not as a Word file to fix.
- in and out as the same file, no Word document in between
- PDF → PDF
- languages in the catalog
- 345
- file formats through the same pipeline
- 13
- surcharge: the words count like any other document
- 0
Layout
The page layout is read, then rebuilt
Before anything is translated, the pipeline recovers the structure of each page: the reading order across columns, the paragraphs, headings and lists, the tables with their cells, and the running headers, footers and page numbers. A tagged file is read from its own tags. Images and borders are kept where they were on the page.
After translation, every piece of text is written back into the box it came from, so the translated page keeps the geometry of the original. Fonts the file does not carry are replaced with metric-compatible faces, so line lengths stay close to the source.
- Multi-column reading order
- Tables, cell by cell
- Lists and headings kept
- Headers, footers, page numbers
- Images and borders in place
- Metric-matched fonts
Text fit
Translations run longer; the type makes room
Finnish or German take more space than English, and a box on the page does not grow. So the translated text scales down to the box the source used, and related text shrinks together: a table stays one size instead of a patchwork of sizes.
"Resize text to fit" is a switch on the Delivery tab, on by default. While you translate, the editor shows the size each row will render at, so a tight box is visible before you download.
- Scales to the source box
- Tables shrink as one
- On by default, switchable
- Size shown while you type
Links and structure
Links and the table of contents stay clickable
Web links, internal links and the table of contents are carried into the translated file, pointing where they pointed before. Bold, italic, underline, superscript and subscript survive the round trip, because formatting is stored as data over the text rather than as tags inside it.
- Clickable table of contents
- Web and internal links
- Bold, italic, underline
- Superscript and subscript
Review and memory
The same memory, terms and checks as every other document
A PDF goes through the same editor as a Word file: translation memory and termbase on every plan, machine translation on models we host, entity protection that masks names, codes and URLs before the model sees them, and live quality checks on every row. A person marks each segment final; nothing is delivered that nobody looked at.
- Translation memory included
- Termbase included
- Entity protection before MT
- Live checks on the row
- A person decides
Quality gate
Checked at delivery, and tested on every change
At delivery, the rebuilt document is compared with the source and structural differences are reported, the same structural QA that guards Word, Excel and PowerPoint deliveries.
Every change to the pipeline is tested against a set of reference PDFs: each is rebuilt and compared with the original page by page for text placement, images and borders, before the change ships.
- Structure compared at delivery
- Reference PDFs rebuilt page by page
- Text placement, images, borders
- Runs on every pipeline change
Plans
Counted like any other document
A PDF costs what any document costs: its source words count against the monthly machine-translation allowance, and there is no surcharge for the format. Every signup starts with 3,000 free words on real documents, Basic has 25,000 words a month at €9.90, and paid plans include a 14-day trial on paid plans.
- No format surcharge
- 3,000 words free to start
- Basic: 25,000 words, €9.90/mo
- 14-day trial on paid plans
FAQ
PDF translation, answered
Upload the PDF to a project, translate it in the CAT editor, and download the result as a PDF. The pipeline recovers the page structure on the way in, the translation is written back into the original layout on the way out, and the type scales to fit where the translation runs longer. You never touch the layout yourself.
A PDF. Converting a PDF to Word and handing back the Word file is a common shortcut, and not one we take: the translated file keeps the columns, tables, images, links and page breaks of the one you uploaded.
The type scales down to fit the box the source text used, and related text shrinks together so a table or a list stays one size. "Resize text to fit" is on by default and can be switched on the Delivery tab, and the editor shows the size each row will render at while you work.
Yes. The pipeline reads the reading order across columns, translates tables cell by cell and keeps lists and headings as lists and headings. A tagged file is read from its own tags; an untagged one is analysed from the page geometry.
No. Machine translation runs on models we host on our own servers in Finland, and entity protection masks names, codes, emails and URLs before the model sees the text. Your file and your translation memory stay on servers we operate in the EU. See EU hosting.
The same as any document: source words count against the monthly machine-translation allowance, with no surcharge for the format. Every signup starts with 3,000 free words, and Basic has 25,000 words a month at €9.90. The full pricing has every plan.
Word, Excel, PowerPoint and OpenDocument files, CSV, RTF, plain text, subtitles and software strings, thirteen document formats in all, each rebuilt to its original format on export. The full list is on file formats.
Try it on your own PDF
Upload a PDF, translate the first page, download the result.