Libris Translate 0.29.0 · User guide
Translate the text of a PDF or scan
Extract a PDF’s text and translate it into a readable book.
Before you start
Have the PDF and its password if required. Your installation must provide an AI model and allow imports. For a scan, identify its language and keep the original to check recognition.
Steps
- In “Add content”, choose “PDF documents” and a series where the document will belong.
- Under “PDF options”, enter the password if needed. Leave “Text recognition language (OCR)” on “Automatic detection”, or enter the scan’s language. OCR means automatically reading letters from an image.
- Drop the PDF on “Files”. Read the extraction warnings shown before confirmation.
- On “Confirmation”, check the languages and estimate, then choose “Import and run the whole pipeline”.
- When the result is ready, export EPUB, Word, text or Markdown. Compare a few sentences with the original PDF, especially after a scan.
What you should see
The PDF’s text appears in a chapter and follows the usual translation workflow. You obtain a readable file in an offered output format, with extraction warnings retained.
What to know
Each PDF becomes one chapter; it is not split at headings. Extraction and OCR run on your server without calling the AI model; translation then sends it text and context. The password is never stored. The original PDF is retained with its protection, and extracted text is saved so an archive restore need not repeat OCR. The licence counts recognised words, not pages. There is no PDF export, and images and page layout are not reproduced. A document is limited to 1,000 pages, two million extracted characters and the installation’s upload limit, 256 MB by default. An XFA form, a type of interactive form, requires a static PDF copy with visible answers. A portfolio, a PDF grouping several files, must be extracted into separate documents. Standard RC4/AES password protection is supported; certificate encryption and AES-GCM are not. Some scans remain unreadable: the warnings help you identify recognition limits.
Fictional example
To translate a three-page archive notebook, the editor imports its English scan. She downloads a French EPUB and checks dates and names against the original document.