OCRmyPDF and Calibre for scanned-document reading

By EPUBTurn · Updated 6 October 2026

OCR can make a scan searchable without making its paragraphs comfortable to read. Verify recognised text first, then inspect the EPUB produced by the next conversion stage.

On this page

Check recognised words before using Calibre
Keep OCR and conversion as two recorded stages
Separate existing scan-text errors from conversion errors
Searchable PDF is only an intermediate step
Compare the actual scan-reading routes on the same pages

Check recognised words before using Calibre

Create an OCR copy while preserving the original scan.
Check several words and numbers against the page image.
Try conversion on a representative section.
Inspect paragraphs and useful pictures in the EPUB.

Keep OCR and conversion as two recorded stages

A basic OCRmyPDF workflow is ocrmypdf inputscan.pdf searchable.pdf. Use separate input and output names and follow the current installation documentation. Open searchable.pdf and compare recognised text with the scan before moving on. The purpose of this stage is a searchable PDF, not a promise of paragraph reflow.

Then add the OCR PDF to Calibre and use Convert books with EPUB as the output for a representative trial. Record any conversion settings you change. Inspect the resulting paragraph order and diagrams independently of the OCR check. A recognised sentence may still be placed in the wrong reading order. If many pages need repairs, weigh that work against an EPUBTurn sample of a permitted source. An online reading conversion is a different workflow and should not be treated as an invisible extension of the local OCR tool.

Separate existing scan-text errors from conversion errors

The cookbook excerpt is an actual historical scan with an existing, imperfect text layer. On the Water Bread page, printed fractions and copied source text differ before a converter receives the PDF. OCRmyPDF can be part of a separately prepared local workflow, but replacing or repeating recognition requires reviewing the resulting text against the printed page. Calibre’s conversion and source recognition are different stages. This guide does not claim that OCRmyPDF was run on the excerpt or that a new recognition pass corrected its fractions.

Searchable PDF is only an intermediate step

Searchable PDF is an intermediate result, not proof of comfortable EPUB reflow. Do not run an unfamiliar command over your only copy. Read the current tool instructions and inspect output before batching files.

Compare the actual scan-reading routes on the same pages

The three-page cookbook comparison used exactly the same unchanged PDF for EPUBTurn and Calibre 8.8.0 defaults. EPUBTurn’s sample has recognised Water Bread ingredients and instructions, including inspected fractions, plus both photographs. The Calibre output keeps scan layers separated and has no recognised recipe paragraphs. That observed difference answers this reading task; it does not show what a separately configured OCRmyPDF-and-Calibre workflow would produce. Keep recognition preparation and conversion results distinct.

Open the actual source PDF · Download the saved EPUBTurn sample.

For your own document, create an EPUBTurn sample from your PDF, then use the reading checks described here.

Actual Water Bread body text from the unchanged three-page input; retained fractions and ingredients. Open full-size image

Actual Calibre 8.8.0 output from the same three-page PDF; separated scan mask rather than recipe body text. Open full-size image

Original Water Bread recipe, printed page 54 in Farmer’s 1917cookbook. — Original Water Bread ingredients and method; both converter outputs above used this same source page. Open full-size image

Files used in this guide

Download the source files, conversion results or worksheets referenced here. Each label identifies the source document or the converter used.

Same three-page original cookbook input Actual unedited EPUBTurn three-page sample