Back to Dashboard

How to Compress a PDF (and When Not To)

PDFCraft's compressor turns every page into a JPEG — great for scans, wrong for contracts. Here's how to tell which one you have.

A 40-page scanned contract that won't fit under your email provider's 25 MB limit, and a 40-page Word-to-PDF export that grew from 600 KB to 4 MB after you "compressed" it, are two different problems. The first is exactly what a PDF compressor is for. The second is what happens when the wrong kind of compressor gets used on the wrong kind of file. Before running any tool on either one, it helps to know why PDFs get large in the first place, and which kind of compression fits your document.

Why PDFs Get So Big

Four things usually drive file size, often together:

Scanned documents are dominated by the first two causes. Born-digital documents — anything exported from Word, Google Docs, InDesign, or similar — are usually dominated by the third and fourth, and their actual page content, real text plus vector graphics, is already extremely compact.

Compressing Images vs. Rasterizing the Whole Page

"Compress the PDF" can mean two very different operations, and the difference matters more than the percentage saved.

Recompressing images in place means finding the embedded photos inside the PDF, downsampling and re-encoding just those image objects, and leaving everything else untouched. Text stays as text, vector lines stay as vector lines, links and form fields keep working. This is what Adobe Acrobat's "Reduce File Size" and similar desktop tools do.

Rasterizing the page means rendering the entire page — text, vector graphics, form fields, everything — to one flat image, then rebuilding the PDF around it. It's effectively the same pipeline as running PDF to Images and feeding the results into an images-to-PDF conversion, done as one step. That's the right technique for a scan that was already just a picture. Applied to a page with real text on it, it throws the text away and replaces it with a picture of the text.

What PDFCraft's Compressor Actually Does

PDFCraft's Compress PDF tool takes the second approach. It renders every page to a canvas at your chosen resolution, re-encodes that canvas as a JPEG, and rebuilds the PDF around the resulting images, page by page, entirely in your browser. There's no partial approach: every page becomes a JPEG, whether it started as a scan or a paragraph of crisp digital text.

For a scanned or photo-heavy PDF, this is close to ideal — the page was already a picture, so a smaller picture loses nothing you had. For a text document, contract, form, or anything that must stay searchable, selectable, or accessible, it's the wrong tool: the text layer is gone, so you can't select or search text, hyperlinks and form fields stop working, bookmarks are dropped, and a screen reader has nothing to read — it sees one image per page, not words.

There's a size consequence too. JPEG is built for smooth photographic gradients, not the sharp black-on-white edges of rendered text. A dense text page rasterized at 150 DPI can easily produce a 150-400 KB JPEG. Multiply that across a document that started out a fraction of that size as real text, and the "compressed" file comes out larger than the original — an inherent property of turning text into pixels, not a bug specific to PDFCraft.

Heads up: whichever level you choose, keep your original file. Open the compressed output afterward and check it still does what you need — try selecting a line of text, clicking a link, or tabbing into a form field. If any of that stopped working, you rasterized a document that needed to stay as text, and the fix is to start over from the original with a different approach, not to keep adjusting the compression level.

Choosing a Compression Level

PDFCraft offers three levels, each mapped to a resolution and a JPEG quality setting:

LevelSettingsBest forRealistic outcome
Low Compression220 DPI, 85% qualityPhoto-heavy pages you may reprintModest savings, roughly 15-35%; sharpness held onto
Recommended150 DPI, 65% qualityStandard scans, mixed text-and-photo pagesSolid savings, roughly 40-70% on scans; still readable
Extreme96 DPI, 40% qualityScans you'll only view on screenLarge savings, 70-90% on scans; visible blur, small print can go blotchy

In real numbers: a 20-page black-and-white scan around 25 MB at 300 DPI typically lands around 4-7 MB at Recommended, under 2 MB at Extreme, usually still legible unless it had small dense footnotes. Run the same settings on a 20-page born-digital contract that started at 500 KB and expect a loss, not a win: all three will likely produce a larger file, because there's little to save and JPEG re-encoding adds overhead text-as-text never carried.

Should You Use This Tool? A Decision Table

Document typeUse this tool?Expected outcome
Scanned paper documentYesLarge size reduction; looks the same at normal zoom
Photo-heavy brochure or portfolioYes — Low or RecommendedModerate reduction; use Low if you'll reprint it
Screenshot-heavy reportCautiouslySome reduction; small on-screen text can blur at Extreme
Word/Google Docs export, resume, born-digital contractNoFile likely gets bigger; selectable text and links are lost
Fillable PDF formNoForm fields are flattened into an image and stop working
Anything needed for e-discovery, OCR, or accessibilityNoText layer and metadata removed, not just shrunk

What to Do Instead for Text-Heavy PDFs

If your large file is a text document, the fix isn't a different compression setting — it's a different approach:

If the bloat comes from leftover objects picked up in an earlier merge, PDFCraft's Merge PDF tool rebuilds a clean document by copying pages into a fresh file with pdf-lib, so orphaned objects do not carry over. It will not fix duplicate fonts, though — pdf-lib does not deduplicate resources across the files being merged, so combining several documents that share a typeface embeds that typeface once per source file, and the result can end up larger than the sum of the originals. See merging and splitting without uploading for how that works. And if you're weighing compression against converting pages to standalone images for another workflow, converting between PDF and images covers the same page-to-picture pipeline a rasterizing compressor uses internally, as a step you control yourself.

Frequently Asked Questions

Why did my PDF get bigger after I compressed it?

Almost certainly because it was a text or vector document, not a scan. Rasterizing compressors like PDFCraft's turn every page into a JPEG, which handles photographic detail well but is inefficient at sharp text edges, so a page that started as a few kilobytes of real text can become a 150-400 KB image. Use a source-app export preset or a desktop tool that recompresses images in place instead.

How small can I realistically get a scanned document?

A typical black-and-white or grayscale scan usually drops 70-90% at the Extreme setting (96 DPI, 40% quality), though small footnotes and fine print will show visible blur at that level. Recommended (150 DPI, 65%) is the safer default for anything you might need to read closely later.

Can I compress a password-protected PDF?

Not directly. Run Unlock PDF first to remove a password you know, then compress the unlocked file. Compression rebuilds the document from scratch, so it can't work on an encrypted file it can't read — and PDFCraft also refuses to re-protect an already-protected file, since that would destroy it.

Will compressing invalidate a digital signature?

Yes. Compression rebuilds the entire PDF around new image objects, changing the underlying file content, so any existing digital signature will show as invalid afterward. Compress a document before it gets signed, not after.

Can I get the selectable text back after I've already compressed a document?

Not from the compressed file itself. Once a page is rasterized, the text only exists as pixels. Recovering it means running OCR on the images to regenerate a text layer — a separate process with its own accuracy limits, not something the compressor offers.

Is there a version of PDFCraft's compressor that skips rasterizing text pages?

Not currently. It always renders full pages to JPEG, which is why this guide recommends a source-app export or a desktop tool such as Acrobat's "Reduce File Size" for any document where the text needs to survive.