Information after an OCR scan is lost - half of the document is missing

There’s a rotation applied in the image that appears to be messing this up.

Orientation : Rotate 90 CW

Development will have to look into whether this is a DEVONthink or ABBYY issue.

This is fixed and will be in next beta

@aedwards - I’m still seeing this problem in DT3b2 - is that known, or do you need more details?

It’s an issue with the ABBYY engine and we are in discussion with them about it. Thanks for your patience and understanding.

1 Like

this is only FYI, because I don’t know whether any changes have been in beta 3 made regarding this problem: the problem is still apparent in beta 3. Although the OCR result is now different (maybe a coincidence), the document is still not complete.

I’m not sure if this specific issue was consistently fixed. I know there are discussions about OCR going on right now though.

I’m new to DEVONthink and got the same problem. After excluding all other causes, it seems to be the same problem on my system.

My workflow looks as follows:

  • Brother ADS-2800W (Auto Start Scan Mode)
  • PDF (no autorotate, no OCR) on a SMB share
  • MacOS folder action on SMB scan folder: “DEVONthink - Import, OCR & Delete”
    up to this point, all is fine

as soon as the document gets imported and OCRd, landscape pages get rotated and cut off on the right side while a white area on the left is added.

Unfortunately I didn’t figure this out during my trial… hopefully this will be fixed soon, as I am now kind of stuck with no idea how to proceed meanwhile.

Welcome @BergNerd

I can’t comment on “soon”, as we don’t give release dates.

  • Are the original scans in your system Trash still?
  • Do you not have the original documents to rescan?

Thanks for the reply, @BLUEFROG.

I still have the originals for re-scanning, so no information is lost. It was my fault not to check the inbox earlier… I recognized this after scanning about 900 documents… :roll_eyes:

Okay, but that is the risk when using beta releases…

My question ist more how to rescan the documents without running in the same problem again… is there any workaround or do I have to wait until it is fixed?

Thank you!

I recognized this after scanning about 900 documents…

Yikes! :flushed:

Okay, but that is the risk when using beta releases…


According to: , the scanner software has OCR capabilities. You could use their OCR then just import the files post-conversion.

A workaround is to rotate the landscape pages first in DT, and then run OCR. That works fine, at least on my setup.

Which of course might not be fun if you have 900 documents and the position of the landscape pages is unpredictable.

Looking forward to a proper fix. The auto-rotate feature is really useful.

I am saw this problem yesterday as well (DT pro 3.0). It is a long strip shaped document scanned with my ScanSnap s1300i. My only option was to scan-to-file and then import without OCR (and stay not-OCRd’). Is there a fix in the works?


Hi - yes, I did see that, but it was written 3 months ago and when DT 3 was in beta. I wasn’t;t sure if you guys thought it had been fixed or not.


We are still talking to ABBYY about this.

After reading this I am insecure. At what kind of workflow and which settings exactly does the problem occur ? The information does not seem to be unimportant.

It happens during OCR if DT/ABBYY needs to rotate any pages automatically (e.g. if your scan is done in portrait mode but some or all of the pages are actually landscape). If you rotate those pages before applying OCR then AFAIK it’s fine.

Still broken. This is a data-loss issue for me. Any news?

This is an issue with v11 of the ABBYY OCR, at present I would recommend either:

  • rotate pages prior to OCR
  • In DEVONthink 3->Preferences in the OCR tab, turn off the auto correct for “Page Orientation”
1 Like

If you have a standalone FineReader 12, it solves the issue without any problems (just tested it). Here I’ve written a script for DEVONthink to use it with a smart rule. You can setup any setting you want, and make a smart rule to do it automatically.