Skip to content

Translate a document and keep its structure

Upload a Word or PDF file and get it back translated, with headings, lists, tables, images and bold or italic words where they were.

Processed by our own AI server. Your text is processed using an AI model hosted on VersoFile's own server. It is not sent to third-party AI providers or retained as a saved document.

Processed by a language model running on VersoFile's own server. No third-party AI service is used, and your document is not stored, logged or used for training; it is deleted with the result.

Up to 12,000 characters (about 4–5 pages) and 10 PDF pages per document. A full document usually takes several minutes because the model runs on our own processors.

or drop it hereSupported formats: DOCX, DOC, PDF

Options

Result file

Up to 30 terms, each up to 80 characters. Links, email addresses and code are always kept.

Headings, lists, tables, images, headers and footers stay in place. Arabic results are set right-to-left.

This tool uploads your file to our server over HTTPS. It is processed in an isolated temporary folder and deleted after download or automatically within about 15 minutes.

AI can make mistakes. Always read the result before using it, especially names, figures and quotations.

How to use Translate Document

  1. 1

    Upload a DOCX, DOC or PDF file.

  2. 2

    Choose the language to translate into, and optionally terms that must stay unchanged.

  3. 3

    Select Translate document, read the report, then download the translated file.

About Translate Document

Translate Document works paragraph by paragraph inside your file instead of copying its text into a new one. Headings, lists, tables (including nested tables), headers, footers, text boxes and footnotes are translated where they are; images, styles, numbering and page setup are left untouched. Bold, italic and linked words are marked so the model can keep them on the matching translated words; when that fails for a paragraph, the translation keeps the paragraph's first style and the report tells you where. Links, email addresses, code and the terms you list are copied unchanged, and numbers are checked after translation. Arabic results are set right-to-left. Text PDFs are first converted to an editable document and scanned PDFs are read with OCR, so their layout is approximated. Translation is done by an open-weight language model running on VersoFile's own server, not by a third-party AI or translation service. Your document is not stored, logged or used for training, and it is deleted with the result. The model suits documents of a few pages; formatting cannot be pixel-perfect because translated text has a different length, so review important documents before you use them.

Frequently asked questions

Is my document sent to Google Translate or another AI company?

No. The model runs on our own server. Your document is not sent to any third-party service, is not stored after processing and is not used for training.

Will the formatting be exactly the same?

Structure and styles are kept, but translated text has a different length, so line and page breaks can move. If formatting inside a paragraph could not be matched, the report lists that paragraph.

Can I translate a scanned PDF?

Yes. Scanned pages are read with OCR first. Recognition errors can affect the translation and the original layout is not kept, which the report states.

How large can the document be?

Up to about 12,000 characters (4–5 pages) and 10 PDF pages. It usually takes a few minutes because the model runs on our own processors.