[Feature Request]: Store ocr images for lookup for performance

[singersbalm]

What would your feature do?

the image gets sliced into speech bubbles/lines to be processed. I suggest to store already detected areas and the text, translation(with translator and language) of each speech bubble. the main reason for that is that I came across a webtoon with a gif in it. the performance could be drastically improved by storing the speech bubbles.