Is there a way to OCR inline image

This is the one feature I find that I really need in DTP.

I capture a lot of screenshots during development and inline them into the notes I’m taking. I’m thinking of the way Evernote and OneNote do it, where the image is searchable, while being presented as an image in the note.

Is there a way to do this?


This isn’t supported yet.

+1 for this request. This would be REALLY handy.

Is this something which is considered?
Or should you automatically convert images to pdf+text using the OCR.

Is there some way to do this automagically? If pasting an image, convert and OCR to pdf?

Well, we consider almost everything but that doesn’t mean that it will be added soon or is already scheduled/planned.

1 Like

Yes ofcourse! Fully understand :-). I just assume I am not the first one thinking of the image OCR recognition question and just wondering what the best workaround is for this?

Some script or conversion method where you have to convert everthing to a pdf + text as I understand now, PDF + text is still the best way to archive things?

Only drawback I see for that method is that a pdf is not shown by default in a note compared to an image.

If the images are stored on their own in the database, then Data > OCR > to comments/annotation might also be an option.

Use case: Taking notes during a presentation. Using screen captures to grab information from the screen and paste them inline with the notes. At the end, I end up with an RTF with 10-20 embedded images.

There is lots of text in the screen captures that would be of high value to me to be searchable.

Thank you!

@Jjason, try using markdown rather than RTF.

  1. First create a Group where the screenshots would go (I collect images I insert into markdown files in a Group called Assets and I have this at the root level of the Database. That way you can always refer to the image as /Assets/<image_name> (note the / before Assets).

  2. Tell Devonthink where to save the images in Preferences > Files:

  3. As you are taking notes in markdown, when you have taken a screenshot and want to insert it, just paste the screenshot directly (CMD-V or right-click > paste etc) at the insertion point.*

  4. Bonus - set up a rule to OCR automatically:

NB: OCR to comments.

  • Note that you also have access to MacOS’ “import from iPhone/iPad” by right-clicking at the insertion point (the devices have to have the same AppleID and on the same WiFi network). I use this feature to insert quick diagram sketches from my iPad.

edit: typos


A further thought on this method.

If you perform a search and it finds a screenshot through the OCR, it will take you to the image file.

Open up the Links inspector to find the original document:

Untitled 2

And this brings us to another advantage of using markdown - you can insert that same screenshot into any number of other documents (say you have reviewed and consolidated your notes into another document).

1 Like

Thank you! Very valuable feedback! It shows how powerful DT is but also how much you need to rely on the forums to get tips like this! Thanks! I more and more start to see the value of using markdown in the notes!

For me working with screenshots is a daily thing. even for example flight info I take a screenshot and then it is handy to just be able to search images. So I hope one day this can go automagically so beginners do not have to setup magic rules!

Any disadvantage to do this for All images in the database using a smart rule? So that automatically all images are OCRd?

I’m not sure that I follow: you have flight info which you do not add to your calendar but instead take a screen shot of that you then OCR? That sounds terribly complicated for a basically very simple task.

Hi I agree, its more that I wanted to indicate that I use a lot of times just a screenshot in my notes to add extra information especially for on the go. Which is not always written down. So the screenshot holds a lot of information which could be useful to search on. So having auto OCR seems very useful but I could be wrong ofcourse!

Would you want this image OCR’d ?




wow, that Paste-Image-to-Markdown-and-Auto-OCR-to-Comments/Annotations-Approach is very impressive - thanks for sharing.

I still prefer to convert them to PDFs because any annotations are non-destructive then (if you have a need for this, like me :slight_smile:).

But, as said above, that inline-images-in-mardown-with-ocr-in-comments-and-referred-links-in-info-panel-approach is pretty cool (and, on the other hand, shows, how easy it was in Evernote - but that‘s a complete different story - and, on the other other hand, in Evernote you don‘t have access to the OCR data, you can just search for it)

Haha, :slight_smile: you made me laugh

Maybe your AI would maybe already add an annotation “Pig” , “Grass”.

I get your point, but generally the screenshots are screen captures of programs, or slides from a teams meeting etc so they hold interesting info sometimes I will only need later on without realising.

I also found out that the Assets folder is per group?

  • I was making a group for a customer.
  • Created a markdown note
  • Pasted an image

Now my image is in an Asset folder in that group?
Not sure how I could apply a smart rule to OCR all the Asset folder files?

I now realise that the image questions is not so easy to solve.
I wanted for example to open the markdown note in the Typora editor but then
the image is not shown as the link is towards

So it agains makes me think how to handle images within notes.
Would be lovely though to have a live render of the markdown but I assume thats something one day could land in DT as a feature.

Glad you enjoyed it (and I love that piglet :heart: )

  • Are you working within a single database?
  • Is DEVONthink open when you’re busy screencapping these things?

Currently not, I have a Work database and a Personal database.
And then I use groups to put structure into my notes (a bit of a folder approach)

Just as a start no full system yet how to do things

But are you using screencaps in both the Work and Personal databases or just say, the Work one?