I’m using version 4.3.2 of Devonthink Pro. However, I’ve suddenly lost access to the OCR functionality. The OCR tab is no longer displayed in the settings. It’s no longer possible to convert a PDF into a searchable PDF. I can no longer select the Abbey add-on. It looks as though my Pro licence isn’t being recognised.
Is it definitely a Pro license? According to the description (no OCR settings) this sounds like a standard license actually.
Yes, it is
What’s the prefix of your license - e.g. DT3-, DTP- or DTS-?
The Prefix is: DTP3-
Does it work again after downloading the app from our website (DEVONtechnologies | Apps) and replacing the installed copy with the downloaded one?
After reinstalling the app, it’s working again. However, during installation I received an error message from another app (Acronis True Image). Acronis’s security feature flagged the “DEVONthink MCP” file as dangerous and blocked it. Once I disabled Acronis’s security feature, the installation completed without any errors. Thank you very much for your help.
Interesting. I’ve been using DTP for my OCR needs for a long time, basically automating by watching a directory, then pulling all PDFs that have word count of 0 in, running the embedded OCR, then filing into “the next directory.” This after I had used Abbyy myself in a docker container for a while (which turned out to be much slower than what DTP’s embedded Abbyy did).
Now, because I’ve started working on my own auto-filing system I ended up finding out that using Apple’s embedded OCR capabilities is actually maybe a factor 10 faster still - I’m currently writing a PDF editor which allows keeping a ledger on decisions you took on a document (think, invoice paid / filed etc.) and all the obvious annotations (think, Skim) whilst also right away embedding OCR as part of the tool. This will replace my DTP based PDF workflow, as the whole automation around that editor will also allow e.g. an agent to auto-file in my repository tree.
If you are referring to Apple’s Vision/Live Text, you are aware this is not doing traditional OCR, therefore no text layer is added to the document… correct?
Correct, which is why I’ve implemented that 