German Umlauts "destroyed" when converting to PDF/A

[docbobo]

I am trying to use OCRmyPDF to convert an existing PDF to PDF/A. Command looks like this

ocrmypdf --output-type pdfa --skip-text input.pdf output.pdf

In the resulting output.pdf German special characters are destroyed. When using pdf (not pdfa) as output type, the special characters are maintained correctly. Am I doing something wrong here?

[jbarlow83]

It should help to specify --language deu (or any other languages that may appear) - otherwise the default is English and there is a bias in OCR to dropping characters/words unlikely to appear in the target language.

Although, that shouldn't cause characters to be dropped on PDF/A.

One thing to try is use --output-type pdf, confirm special characters appear, then try rewriting the PDF using
gs -q -sDEVICE=pdfwrite -o out.pdf in.pdf. This command should reconstruct an equivalent PDF. If that rewritten version drops characters, the issue may lie with Ghostscript, which handles PDF/A.

[tcurdt]

That's a litte confusing. I thought OCRmyPDF just adds a layer?
Why would it change existing characters, @jbarlow83 ?

[jbarlow83]

It's also possible one of the blacklisted Ghostscript versions was used (10.0.0 through 10.3.0 inclusive). These are known to corrupt output.

Without a reproducible test case and information about versions used, I can only an educated guess.