Sometimes I have a series of screenshots of a non-English journal article from DeepDyve which I want to (1) translate into English and (2) store in searchable form on DTP.
For a German article from the 1970s, I had better success first letting Google translate the screenshot (.png format) into English (still in .png format), then let DTP use its OCR magic on the English .png to produce a searchable PDF.
I first tried letting DTP use OCR to produce a searchable German PDF from the .png screenshot, then have Google translate that file. The translation results were mainly gobbledygook, for some reason.
I’m not especially partial to Google Translate, as I’ve used DeepL with good or better results. But I’ve used up my free access to DeepL for the time being.
A really robust way I’ve been doing translations lately is to convert to an image (like PNG) and import the page into Apple’s Photos app. Then I select the text and hit the “translate” button. This has changed slightly on Sequoia I see, it’s now even better. The advantage to this approach is that Apple appears to be using some sort of AI lite to make proper words in context. In other words, you don’t get the OCR gremlins that are usually in the text layer. The other thing I do is copy and paste the text from Photos into an annotation in DTP so I have the good text layer as well, associated with my document. This allows me to search on terms and make edits where necessary to improve the search results. (adding diacritics or whatever is necessary after Photos is done).
The only disadvantage at the moment is that a rather small subset of the world’s languages are currently supported. For those that aren’t, I imagine you could take your nice text selection and pop it into some other translation tool.
I hope the attached screenshots give you an idea of what is possible.
Hi, again, wotiva. I tried your approach several times and can see how it could produce good results. With the German journal article I experimented with, the translation results were better with Google Translate. What I’ll continue using of your suggestion, however, is using Apple Photos to enhance fuzzy PDFs. Several of my PDFs are pretty low quality, but the so-called AI button on Apple Photos really sharpened them up. I suspect that’s something that many photo apps could do, and have been capable of doing for years, but I’d not thought to try that. So, thanks!
