The damage cannot be undone. Compression removed pixels and detail, and nothing recreates them. Go back to the original and compress again more gently.
If the original is gone, the only real options are rescanning, regenerating from the source document, or accepting the file as it is.
The file went from 40MB to 3MB and the text now has soft grey edges. That trade was made by whichever setting you chose, and it is worth understanding precisely what happened, because the same understanding tells you how to avoid it.
What compression removed
Two separate operations, both lossy, both permanent.
Downsampling: fewer pixels
A 300 DPI scan reduced to 100 DPI keeps roughly one pixel in nine. The rest are averaged away and gone. This is usually the larger cause of visible blur.
Re-encoding: coarser detail
Each image is decoded and written back at a lower JPEG quality. The encoder discards fine detail it judges less noticeable, which is a reasonable judgement for photographs and a poor one for text.
Detail removed by averaging. There is nothing left to reconstruct from.
Sharpening increases contrast at edges. It does not restore missing detail.
Recompress from the source at a gentler setting.
Why text suffers most
JPEG compression was designed for photographs, where colour and brightness change gradually across an image. It is very good at that, and text is the opposite case.
A letter is a hard black shape against a hard white background: the sharpest possible transition. The encoder handles abrupt transitions by approximating them, which leaves a faint grey halo around every character. On large text it is barely noticeable. On eight point footnotes it is the difference between reading and squinting.
What you can actually do
Ordered by how well they work.
Straightforward
- Compress again from the source, one setting gentler
- Or set a target size rather than a quality preset
- Or remove blank pages first, which costs no quality
- Check the smallest text at 100 percent before accepting
- Never overwrite the original with the result
Limited
- Rescan the paper document if you have it
- Re-export from the source document if it exists
- Ask the sender for a fresh copy
- Run OCR so at least the text is searchable
- Accept the file if it is legible enough
Compressing without this happening
The settings that cause this are predictable, which makes them avoidable.
- Will be printed300 DPI minimum
- Read on screen only150 DPI is ample
- Small text on the pageStay at 300 DPI
- Photographs, no fine textStrong is safe
- Archival copyDo not compress
- Legal or medical recordLight, keep the original
Never compress twice
If a compressed file is still too large, the instinct is to run it through again. That produces the worst results of anything in this area.
The second pass sees the artifacts from the first as genuine image detail and works to preserve them, while discarding more of what remains. You lose noticeable quality for very little additional saving. Always return to the original and compress harder in a single pass instead.
Compress without wrecking it
Light or strong, in your browser, with the original left untouched so you can try again.
Common questions
Recovering and avoiding it.
01Can I undo the compression?
No. Compression permanently discards pixels and detail, and nothing reconstructs them. The only route back is the original file, a rescan, or regenerating the PDF from the source document.
02Will sharpening fix a blurry compressed PDF?
No, and it usually makes it worse. Sharpening increases contrast at edges to imply detail that is not there, which exaggerates compression artifacts into hard halos around soft letters.
03Why is the text blurry when the pictures look fine?
JPEG compression is built for photographs, where tones change gradually. Text is hard black against hard white, the sharpest transition there is, and the encoder approximates it by leaving a grey halo around every character.
04Why did my colleague's PDF compress without going blurry?
Almost certainly because theirs was born-digital rather than scanned. In an exported document the text is vector outlines that compression never touches. In a scan the text is part of the image and degrades with it.
05What setting should I have used?
Match it to the destination. Keep 300 DPI for anything that will be printed or has small text. 150 DPI is ample for screen reading. If the document is archival, do not compress it at all.
06The file is still too big. Should I compress it again?
No. The second pass treats the first pass's artifacts as detail worth keeping and discards more real content, so quality drops noticeably for very little saving. Return to the original and compress harder in one pass.