How to edit a scanned PDF (OCR first, then retype in a matched font)

Someone sends you a scanned contract with a wrong date, or a photographed form with an old address, and asks you to "just fix it". You open the PDF, try to select the text — and nothing selects, because there is no text. This guide explains how to edit a scanned PDF properly: run OCR, retype the block, and export a file that looks like the original and is searchable.

Why a scan cannot be edited directly

A scanner produces an image. When that image is saved as a PDF, the PDF contains one big picture per page and nothing else: no characters, no fonts, no words. A PDF editor that "edits" such a page without OCR can only draw on top of it — usually by painting a white rectangle over the old text and typing new text on top, which is exactly the patchy look everyone recognizes.

To edit a scan like a document, the pixels first have to be turned back into text. That is optical character recognition, and it is the step that decides how good the result can be.

What OCR does and does not do

OCR reads the image and produces text with positions: which words are where, how large they are, and in what reading order. Modern AI OCR also recognizes structure — paragraphs, headings, table cells — and estimates the visual properties of each block: approximate typeface classification, weight, size and color.

What OCR does not do is recover the original font file. A scan does not contain it. So "same font" on a scan always means "the closest matching font", chosen from the editor's library. On clean office documents at reading size, a good match is indistinguishable; on a decorative headline, expect to review the choice.

Scan quality checklist

OCR accuracy depends far more on the input than on the software. Before you upload:

• 300 dpi or better. Under 150 dpi, thin strokes and punctuation blur into each other.
• Straight and flat. Skewed pages stretch letters; photographed pages add perspective. Use a scanner or a document-scanning app that flattens the page.
• Even lighting. A phone photo with a shadow band loses whatever is in the shadow.
• Full page in frame, with a margin, so nothing at the edge is cropped.
• Printed text. Handwriting is recognized best-effort; plan to proofread it.

A clean 300 dpi scan of a typed page usually comes back with essentially no errors.

Step by step with blupdf

1. Upload the scan. Drop the PDF (or a JPG/PNG photo, which blupdf wraps into a page) onto your dashboard. blupdf detects that the pages are image-only.
2. OCR runs on each scanned page. Every word becomes a block with a position, size, color and an estimated font. This happens once per page, at 80 credits per page.
3. Click the block you want to change. It gets a thin outline and the toolbar shows the matched font.
4. Retype. Your new text re-wraps inside the block's original box in the matched font, size and color, so the line sits where the old one sat.
5. Export a searchable PDF. The exported file keeps the original page image, adds a real text layer for every recognized word, and renders your edited blocks as text in the matched font.

Fixing misread characters

Even good OCR trips on a few classic pairs: 0 and O, 1, l and I, rn and m, 5 and S. Because recognized text is just text in blocks, fixing a misread is the same as any other edit — click the block and correct it. Search the document for suspicious patterns (a lone l in a number, for instance) before you export.

If OCR missed a region entirely — a rotated stamp, a faint fax header — add a new text block by hand and position it over the area.

Fonts on scans: closest match, and how to pick a better one

blupdf classifies each block's glyph shapes (serif or sans, weight, width, x-height, spacing) and picks the nearest font in its library. If you know the document was typeset in a particular face, open the block toolbar and choose it from the list; blupdf re-renders the block on the spot. A five-second check at 200% zoom on one edited line tells you whether the match holds up.

Cost in credits

For a scanned document the credits are: 80 per page for OCR (once), 1 per character of new text, and 20 per exported page. A five-page scan with 300 edited characters comes to 800 credits — well within the 1,000 free credits every account starts with. The full table is on the pricing page; Editing scanned PDFs with OCR has the detailed docs.

FAQ

Can I edit a photo of a document, not a scan?

Yes. Upload the JPG or PNG; blupdf wraps it into a PDF page and runs OCR on it. Crop to the page and keep the camera parallel to the paper for the best result.

Will the edited scan still look scanned?

The untouched parts keep the original scanned appearance, with an invisible text layer behind them. Edited blocks are rendered as clean text in the matched font and color, sized to match the surrounding words.

Does OCR work in languages other than English?

blupdf's OCR handles most Latin-script languages and common non-Latin scripts, including mixed-language pages.