How to Compress a Scanned PDF and Keep the Text Crisp

A scanned PDF is an image file that happens to wrap pages. Shrinking it means shrinking the images, and that is where the trade lives: push too hard and the recovered text blurs into pixels, push too lightly and the file stays too large to email or store. This guide is the honest version of that trade — where the lever is, where the stopping point is, and how to check you have not crossed it.

The Rule: Compress the Source Images First

The lever on a scan is never the PDF container; it is the images inside it. So the order is the opposite of what most people try:

1. compress an image the source images — in your browser, free — before rebuilding anything. This is where the size actually comes down, and it is also where the quality decision is made, on the image itself where you can see it.
2. Rebuild the PDF around the compressed images with compress a PDF — in your browser, free, no upload — which then re-packs the container losslessly on top.

Reversing that order — compressing the PDF and hoping the images inside shrink — is why so many people end up with a file that is barely smaller, or text that is suddenly unreadable at the zoom anyone actually reads at.

Find the Stopping Point

The stopping point is a test, not a number: the text is still crisp on the screen the file will actually be read on. A phone preview, a printed page, a 200% zoom in a reader — pick the worst of the contexts the file will face, apply the compression, and look at the smallest text on the page. Keep strengthening until that text stops being crisp, then step back one level and stop. That point is the honest floor for that document; anything past it is quality you will not get back.

What Not to Do

• Do not re-rasterize the whole page. Re-encoding page content as a new image is how crisp vector text becomes a blurry photograph. Compress the embedded images; leave the text alone.
• Do not push past crisp to hit a size target. When a target is genuinely fixed, the target-size method reaches for the lossless pass first and spends image quality in measured increments — rather than sacrificing the whole page at once.
• Do not compress before OCR. If the scan needs a text layer, recognize first, compress after; pixels thrown away before recognition are text thrown away.

When the File Is Still Too Large

If the compressed result is still over the gate, the remaining options are honest and structural: split a PDF the document and handle the parts, or — when the scan has to be searchable too — walk the OCR sequence first, then compress with this guide’s order.

Frequently Asked Questions

Why is my scan still huge after compression? Because the images inside were already compressed, or the compression was applied to the container instead of the images. The source-image step is the one that moves the size; the container step is nearly free on a scan.

At what point does the text get blurry? At the point where the image detail that carries the letter shapes is discarded — which is why the stopping-point test looks at the smallest text, and why it is checked at the zoom people actually use rather than at a thumbnail.

Should I just use the strongest level and get it over with? Only if you do not plan to read the file again. Strongest is a setting for archives that will never be opened, not for documents someone has to read.

---

Compressing a scan is the source images first, the container second, and a stopping point decided by the smallest text at the worst zoom anyone will read it at. Hit that order and the file shrinks without the words dissolving — and the whole job runs in the browser, so the scan never leaves the device it was stored on.

Do It Directly

Ready to shrink the scan in front of you? compress an image the source images — in your browser, free — then rebuild with compress a PDF, no upload. Still over the gate? split a PDF the parts apart, and when the scan also has to be searchable, OCR first (OCR a PDF, on-device, free) before compressing.

Related reading

• Compress a PDF Without Uploading It — the flagship guide: structural recompression, what it recovers, and where the image-heavy limit sits
• Build a Searchable Archive — the OCR sequence this guide compresses the output of
• Free vs Paid on PDFLean — what the local compression path costs and where plan quota starts