Something seriously wrong with new OCR Engine on M1

New OCR engine on M1 Mac makes resulting PDFs blurry.

I have attached three files. Original image, OCR result for M1 and OCR result for Intel.

As you can see image quality of M1 PDF is terrible compared to Intel.

M1 MacBook OCR.PNG.pdf (39.0 KB)
Intel iMac OCR.pdf (324.2 KB)

Do you use the same Preferences > OCR settings on both machines?

yeah, only difference is I have “PDF resolution as source” on M1 but there is no such option on Intel, hence I’ve setted it as 300 dpi. Screenshots attached.

It looks to be an over zealous compression by ABBYY on generating the PDF, if you unselect “Compress PDF”, this should workaround the issue until we release a fix.

Your suggestion solved the issue. Thanks a lot for that.

But now I noticed another weird problem. When I look at “Concordance” inspector, I see different word counts on M1 and Intel Macs for the exactly same file.

To be clear; I’ve downloaded the file titled “Intel iMac OCR.pdf” attached to original post, and looked at the word counts on concordance inspector on both Macs.

Intel DEVONthink shows 182, whereas M1 DEVONthink shows 171 word counts. (You can test yourself by downloading the file and viewing it in M1 and Intel computers)

I think there is another serious bug here.

What operating system is running on each Mac?

both is latest developer beta of Ventura: Version 13.0 Beta (22A5352e)

While it may not be specific to this issue, I wouldn’t expect the best behavior from the beta operating system. It’s definitely still buggy. Have you tried reimporting the file?

1 Like

yeah, I tried reimporting but nothing has changed.

I don’t think that this is beta OS related issue, because both machines have exactly same OS.

But even if it is OS related, take this as an early warning for a bug.

The Intel and Apple Silicon versions of ABBYY OCR are two different releases and as such may exhibit different behaviours. Word recognition is never 100% accurate and it is possible to get differences between versions, however overall accuracy in both version should be over 90%.

No, I am not talking about different results between machines. I am saying that, when I look at the concordance inspector I see different word counts for the exactly same file.

I am not seeing that here. The word counts are exactly the same when I view them on an M1 and Intel Mac (running Monterey) for the “Intel iMac OCR.pdf” file.

1 Like

Then it must be a Ventura problem. Thanks a lot for testing.

1 Like