Turn a scanned PDF into an editable one

How this works, and the trade-offs: each page is rendered and OCR'd on-device, optionally cleaned up with AI, then rebuilt with the recognized text drawn as real, visible text objects — not an image — so PDF editors can select and edit it directly. With "Keep logos, signatures & images" on, the original scanned page is kept as a background image and only the recognized text lines are erased (painted over, assumed to be on a plain white/light background — this can look slightly off on colored letterhead or unusual backgrounds) and replaced with real text in place; everything else — logos, signatures, photos, table borders — stays exactly as scanned, at the cost of a larger file. With it off, the page is rebuilt as text only, dropping everything non-text for the smallest possible file. Either way, only the reconstructed text's approximate layout is preserved, not exact fonts/spacing, and it works best on plain text documents (letters, memos, simple reports) rather than heavily designed pages. OCR accuracy is never perfect, especially on messy scans or unusual fonts.