RetainPDF Tutorial: How to Translate Scanned and Image-Based PDFs

Scanned PDFs are the hardest documents to translate: there is no text layer, formulas are images, and OCR output usually destroys the page layout. This tutorial shows how to get a clean, layout-preserving translation with RetainPDF — from a raw scan to a finished bilingual-style PDF.

Step 1 — Get RetainPDF

Grab the release for your OS from GitHub Releases (Windows Setup.exe, macOS .dmg, Linux .deb), or deploy the Docker image wxyhgk/retainpdf-app:latest if you want a shared instance for your team.

Step 2 — OCR handles the scan

RetainPDF runs OCR on image-based pages automatically. The scan below is converted into a real text layer — the first step toward an accurate translation, not a garbled blob:

Step 3 — Protect code and formulas

Configure translation rules so code blocks stay untouched and complex inline formulas render correctly. RetainPDF's rule-based strategy means you decide what gets translated and what stays verbatim — a lifesaver for technical manuals and research papers.

Step 4 — Export with layout intact

The result keeps the original structure: headings, tables, and page flow survive, and the output is compressed to keep file size reasonable. Here is the scanned-document result after translation:

Automate the whole flow

RetainPDF is decoupled (OCR → translation → typesetting → delivery), so you can call its API and plug it into a document pipeline that processes hundreds of PDFs without human interaction.