Lingivio

Translate PDF and Keep the Images

Images, charts and background graphics are not moved and not replaced. Only text nodes are touched, and on a scan the artwork under a replaced label is rebuilt rather than blanked.

What Survives When You Translate PDF

  • Output is produced by removing the original text and drawing the translation back into the same box, so page geometry, images and vector art are the originals.
  • A dual mode puts the original page beside the translated one, which is what you want when you are checking a document rather than replacing it.
  • Born-digital pages are read from the text layer; scanned pages are routed to OCR automatically, page by page, inside the same file. Text-layer pages are read exactly. OCR pages are best-effort — a line the recogniser cannot read stays in the source language — and they are billed in full either way.
  • A block we cannot translate confidently keeps its original pixels: visibly untranslated, never filled with a guess.

See It on a Real Document

The before and after below is a real PDF file this engine translated (Spanish to English). The examples page has the files themselves.

Source documentENISA Loan Application Guide
Page
Original · ESTranslated · EN
Source: public applicant guide published by ENISA (Spain). Translated by Lingivio, Spanish to English, no manual clean-up.
Download the files on the examples page

What This Does Not Do

  • Scanned pages and images are best-effort: on dense or degraded scans a share of lines is never extracted at all. Those lines stay in the source language on the page rather than being guessed at, and the page is charged in full whatever OCR reads off it. We do not promise a scan comes back 100% translated. Born-digital pages carry no such caveat.
  • Handwriting-style Nastaliq — Urdu in particular — is not read reliably by any OCR we have tested. Those files fail with an explicit error instead of returning something that merely looks translated.
  • Anything the verify stage cannot confirm falls back to the source text, visible and untranslated. A file that looks finished but is not is the one outcome we will not ship.

Questions

Do the Images Survive Translating a PDF?
Images, charts and background graphics are not moved and not replaced. Only text nodes are touched, and on a scan the artwork under a replaced label is rebuilt rather than blanked.
What happens to my file afterwards?
Uploaded and generated files are automatically deleted within 24 hours, and are never used to train a model.

Try It on Your Own File

No card and no per-file character minimum. If a segment cannot be translated confidently you will see it fall back, rather than a green tick over a guess.

Related Pages

Cookies, analytics and diagnostics

Only strictly necessary cookies are used to sign you in and remember your preferences. We would also like your permission for analytics, so we can see which parts of the product are used, and for diagnostics, so browser errors are reported to us. Nothing non-essential runs unless you say yes, and you can change this at any time.Read the Cookie Policy