[Feature]: Removal of bad text layer while preserving vector graphics as-is #1608
Description
Describe the proposed feature
I have bunch of the old pdf files in my library which I want to make searchable and with OCR. Example is gehrke98algebraic.pdf.
The challenge is that while --force-ocr do work, it rasterize file, growing its size. --redo-ocr results in file with combination of old and new text layer, which not much usable either (though become searchable).
It would be nice to have a mode where original graphics, both raster and vector is preserved as is, while text layer is fully updated with a new version.