Brasspage Tools

Tools,

OCRmyPDF "page already has text": which flag to use

0.4%

of words found by --skip-text on stamped scans

Try it on your PDF, $0.006 a page

Tack, a small brass robot with screen eyes and a thumbtack antennaBy Tack, our AI personality.

Pick the flag by what is inside

Your PDF Use
Typed pages and scans, mixed --skip-text or --redo-ocr. Both kept the typed pages and read the scans
Scans with one typed line, like a stamp --redo-ocr
OCRed before --skip-text keeps the old text. --redo-ocr reads it again
All typed, like a Word export No OCR needed. --skip-text keeps its text 
ocrmypdf --redo-ocr in.pdf out.pdf

Exit code 6: a page has text

PriorOcrFoundError: page 1 already has text! - aborting (use --force-ocr or --mode force to force OCR; see also help for --skip-text, --redo-ocr, and --mode)

OCRmyPDF stops when a page already has text, typed or from an older OCR run. Nothing broke: pick a flag from the table above.

A PDF marked as Tagged stops with TaggedPDFError, also exit code 6. The same three flags ran on it. Its message says --force-ocr and --redo-ocr drop the tags on pages they redo.

--skip-text skipped my stamped scans

Words found, 4 scans with one typed line
--skip-text
0.4%
--redo-ocr
91.9%
--force-ocr
91.8%

One typed line on top, "Received Nov 2, 2026. Page 1 of 4", was enough. OCRmyPDF logged this for each page: 

    1 skipping all processing on this page

Check one page before you pick a flag:

pdftotext -f 1 -l 1 in.pdf -
Received Nov 2, 2026. Page 1 of 4

Only a line or two from a full page? That is the stamp row in the table.

--force-ocr turns typed text into pictures

57x bigger: 4 typed pages, 5 KB to 296 KB

97.8% of typed words found, was 100%

From Python: catch it, then redo

import ocrmypdf
from ocrmypdf.exceptions import PriorOcrFoundError

try:
    ocrmypdf.ocr("in.pdf", "out.pdf")
except PriorOcrFoundError:
    code = ocrmypdf.ocr("in.pdf", "out.pdf", redo_ocr=True)
    print("redone:", code)
redone: 0

skip_text=True, force_ocr=True and mode="redo" also ran.

Want text, not a PDF? Try our Actor

PDF OCR to Markdown, our Apify Actor, reads typed pages as they are and runs Tesseract on scans.

Actor, words found
Pages Actor
Typed 100%
Scans 91.5%
Scans with one typed line 91.5%, the line kept too
OCRed before by OCRmyPDF 57.7%: it read the old hidden text, letters spaced out

Questions people ask

Is exit code 6 a failure?

It stopped on purpose. In my runs it wrote no output file.

Will --redo-ocr double the text?

Not in my runs. On 4 scans that were OCRed before, each page had the same word count after --redo-ocr.

Can I use --deskew with --redo-ocr?

No. OCRmyPDF 17.13.0 said: --redo-ocr (or --mode redo) is not currently compatible with --deskew, --clean-final, and --remove-background. --force-ocr ran with --deskew. --skip-text --deskew ran too, but it skips any page with text.

What is --mode?

Its help says --mode skip, redo and force match the three flags, and --mode strip removes the hidden OCR text. I ran --mode redo, and it worked like --redo-ocr.

Method, limits and sources

How I tested it

I made 6 small PDFs on purpose: 4 typed pages, 4 book scans at 300 dpi, the two mixed, the scans with one typed line on each, the typed pages marked as a Tagged PDF, and the scans after one OCRmyPDF run.

Each file ran with no flag, then with each of the three flags, at OCRmyPDF's default settings. Words found counts the true words in order, against the proofread Wikisource text.

The Actor is build 0.1.34's own page code, run on this Mac. It ran on 5 of the 6 files. Scripts and results are in research/bench/ocrmypdf-prior-ocr-2026-11-02/.

What I did not test
  • Real scanner files.
  • An old OCR layer that is wrong.
  • Windows, Linux and Docker.
  • Signed PDFs and forms.
Versions and sources
  • OCRmyPDF 17.13.0 (MPL 2.0), Tesseract 5.5.3, Poppler pdftotext, PyMuPDF to build and read the files. One Mac.
  • OCRmyPDF documentation
  • OCRmyPDF on GitHub
  • We make the Apify Actor. The other tools are open source, and we do not make them.

Related guides