OCR with user selecting the part of the photo that has the target text instead of automatic selection

Before uploading the photo, allow the user to select a cut. He manually selects the area where the text is for OCR and it is this area that is loaded and converted to text. In this way, a single photo could be used for several columns and there is no need to store several of them as it would be enough to have a metadata for each column storing the four guide points of the selected area.