Compress PDF Without Losing Quality
· 5 min read
There are two methods and they are not variations of the same thing. Structural compression rewrites the file more efficiently and keeps text selectable, saving 5-20%. Rasterising renders each page to an image and removes 60-80% from a scan, but the output has pictures of words rather than words, and it cannot be reversed.
There are two ways to shrink a PDF and they are not variations of the same thing. One is safe and modest; the other is dramatic and irreversible.
On this page
Structural compression keeps everything
The safe method rewrites the file more efficiently: packing objects into compressed streams, removing unused metadata and deduplicating repeated resources. Nothing about the content changes, so text stays selectable and searchable.
The saving is real but modest — often 5–20%, occasionally more on a badly-exported file. If your PDF is already lean, it may be almost nothing, because there is nothing wasteful left to remove.
This is the mode called Keep text selectable in FileCrisp. Its promise is narrow and absolute: the document you get back is the same document, just stored more tidily.
Rasterising is where the large savings live
The aggressive method renders each page to an image and rebuilds the document from those images. On a scan this routinely removes 60–80%, because the embedded photographs are re-encoded at a sensible resolution instead of whatever the scanner produced.
The cost is that the output contains pictures of words rather than words. Text can no longer be selected, searched or copied, and no tool can turn it back. Keep your original.
Both Balanced and Smallest file work this way. They differ only in resolution — 150 dpi against 110 — which is the difference between clean printed output and softness that shows on paper.
Compress PDFCut PDF size for email without wrecking the scanThe trade-off, stated plainly
Most compression tools present their modes as a simple quality slider, which hides the only decision that actually matters. These two methods are different operations with different consequences.
| Structural | Rasterising | |
|---|---|---|
| Typical saving on a scan | 3–5% | 87–93% |
| Typical saving on a text PDF | 5–20% | Large, but destructive |
| Text stays selectable | Yes | No |
| Links and form fields survive | Yes | No |
| Reversible | Nothing to reverse | No |
| Safe for a document you will search | Yes | No |
Which one you want
If the document is a scan, rasterise. The text was already an image, so you lose nothing that existed and gain most of the file size back.
If the document has real text you or the recipient will search or copy, use the structural method and accept the smaller saving. If that is not enough, split the file rather than destroying its text.
- Scan or photo-heavy PDF → rasterise, expect 60–80%
- Text PDF you need searchable → structural only
- Fillable form → structural only, always
- Text PDF still too big → split it and send in parts
Why 'without losing quality' is usually the wrong goal
Losslessness is only worth paying for where it changes an outcome. For a scanned receipt that will be glanced at once and filed, insisting on lossless compression means keeping a 22 MB file to protect detail nobody will ever look at.
The better question is what the document has to survive: being searched, being printed at full size, being read on a phone. Answer that, and the mode chooses itself.
How to check you made the right choice
Open the output and try to select a paragraph. If it highlights, the text survived. If nothing selects, you rasterised — which is correct for a scan and a mistake for a contract.
Then look at one page at 100% zoom. Compression damage never shows in a thumbnail, and a thumbnail is not how the recipient will read it.
Common questions
Can you compress a PDF without losing quality?
Yes, with structural compression, which repacks the file's internal streams and drops unused metadata without touching page content. Text, links and fonts are preserved exactly. The saving is modest — typically 5–20%, and near zero on a file that was already efficiently written.
Which compression mode should I use?
Keep text selectable for contracts, reports and anything fillable or searchable. Balanced for scans, where the text was never real text to begin with. Smallest file only when a hard size limit leaves no choice, and check the result before sending.
Does compressing a PDF affect print quality?
Structural compression does not — the printed output is identical. Balanced renders at 150 dpi, which still prints cleanly at A4. Smallest file renders at 110 dpi, where softness becomes visible on paper even though the screen view looks fine.
Why is my PDF still large after compression?
If you used the lossless mode on a scan, there was little for it to do — the size is in the page images, which it deliberately does not touch. Switch to Balanced. If you already did and it is still too large, the document is simply big; split it.
Can rasterised text be turned back into real text?
Not by undoing the compression — the words are gone and only pixels remain. OCR can produce new text by recognising the image, but that is a fresh guess at the content rather than a recovery of the original.
Does structural compression change how the document looks?
No. It changes how the file is written, not what it draws. Every page renders identically before and after, which is precisely why the savings are limited.
Is one pass enough, or should I compress twice?
One pass. Running a rasterising mode twice re-encodes images that are already lossy, adding damage for almost no further saving. If one pass did not get you there, change the approach rather than repeating it.