How to compress a PDF file for free

By Nasir Uddin Shamim · Founder, PrintxPDF

Updated 2026-08-18 · 7 min read

Short answer: To compress a PDF for free, open it in a browser-based compressor and pick one of two methods. Lossless structural compression subsets fonts and strips unused objects without touching a pixel. Image downsampling re-encodes photos at a lower DPI. Text documents shrink 10-30 percent; scans can drop by 90.

Every PDF compressor on the internet does one of two things, or both. Knowing which one your file needs is the whole game, because applying the wrong one either does nothing or ruins your document.

The two kinds of PDF compression

Structural (lossless) compression rewrites the plumbing of the file. It packs small objects into object streams, replaces the cross-reference table with a compressed cross-reference stream, deletes objects nothing points to any more, and subsets embedded fonts so only the glyphs you actually used are stored. Nothing you can see changes. Every pixel and every character survives exactly.

Image (lossy) compression re-encodes the pictures inside the PDF. It downsamples them to a target resolution and pushes them through JPEG at a chosen quality. This is where the big numbers come from, and it is a one-way door: the discarded pixels are gone.

Method	What it changes	Typical saving	Reversible
Font subsetting	Stores only the glyphs used	50 KB - 8 MB per font	Yes, visually identical
Object streams + xref streams	Packs metadata objects together	5-15 percent	Yes
Removing orphaned objects	Deletes unreferenced data left by edits	0-40 percent	Yes
Image downsampling	Reduces pixel dimensions to a target DPI	60-95 percent on scans	No
JPEG re-encoding	Lowers image quality to 60-80	30-70 percent on photos	No
Flattening layers and forms	Merges annotations into the page	5-20 percent	No

What each method does to a typical file.

Compress a PDF in your browser

Open the compressor Go to the PDF compressor and drop your file on the page. Nothing is uploaded, the file is processed by your own browser, so a confidential contract never leaves your machine.
Try lossless first Run the structural pass on its own and look at the result. If a 4 MB text document drops to 2.6 MB, stop there. You have paid nothing in quality.
Check whether images are the problem If lossless barely moved the needle, the weight is in images. Scroll the document: if the pages look like photographs of paper rather than crisp text, it is a scan and image compression is the only lever that matters.
Pick a level Three choices, not a DPI box. Light is lossless: it rewrites the file structure and drops metadata, keeping text selectable. Medium re-renders each page at roughly 100 DPI and re-encodes it as JPEG. Strong goes to about 72 DPI. Medium and Strong are what shrink scans, at the cost of turning text into pixels.
Compare before you commit Open the compressed file and zoom to 200 percent on the smallest text on the page. Compression artefacts show up around letter edges before they show anywhere else.
Keep the original Save the compressed copy under a new name. You cannot un-downsample an image, so the original is your only fallback.

Compress a PDF in your browser. No upload, no account, no watermark.

Choosing a resolution

Target DPI	Pixel size (A4)	Approx. size per page	Use for
600	4960 × 7016	3-8 MB	Archival masters, fine artwork, OCR of tiny print
300	2480 × 3508	800 KB - 2 MB	Anything that will be printed
200	1654 × 2339	400-900 KB	Good compromise for scanned contracts
150	1240 × 1754	200-500 KB	Screen reading, email, most business documents
96	794 × 1123	90-200 KB	Web preview, thumbnails
72	595 × 842	60-150 KB	On-screen only, never print this

DPI targets and what they are for. Sizes are per A4 page of a colour scan.

Note: 300 DPI is the standard for printing because it is roughly the point at which the human eye stops resolving individual dots at normal reading distance. Going above it for a document that will only be emailed is pure waste.

What lossless compression actually removes

Unused font glyphs. A full CJK or large OpenType font can be 5-20 MB. If your document uses 40 characters of it, subsetting stores 40 glyphs.
Orphaned objects. Editing tools frequently write new objects and leave the old ones behind. Nothing references them, but they still occupy bytes.
Old revisions. PDF supports incremental saving, where each save appends to the end of the file instead of rewriting it. A file saved thirty times can carry twenty-nine dead copies of its earlier self.
Duplicate resources. The same logo embedded once per page becomes one shared object.
Uncompressed streams. Some generators write page content streams with no Flate compression at all. Simply zipping them can halve a file.
Bloated metadata. XMP packets, thumbnails and structure trees left behind by design tools.

When compression does almost nothing

A clean, text-only PDF exported straight from Word or LaTeX is already near-optimal. Expect 5-15 percent, not 80. If a compressor promises to shrink a 900 KB text PDF to 90 KB, it is rasterising your pages into images, which destroys the text layer and makes the file unsearchable. Check afterwards: press Ctrl+F and search for a word you can see. If nothing is found, your text layer is gone.

Watch out: Never compress a signed PDF. Any change to the bytes invalidates the digital signature. Compress first, then sign the PDF, never the other way around.

Compressing on the desktop

If you prefer local tools, three work well:

macOS Preview, File → Export, then set Quartz Filter to "Reduce File Size". It is aggressive and slightly crude, but it is already installed.
Adobe Acrobat Pro, File → Save as Other → Reduced Size PDF, or Tools → Optimize PDF for per-image-type control over downsampling.
Ghostscript, free, scriptable and the engine behind many online compressors.

# 150 DPI, good for email and screen reading
gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.7 \
   -dPDFSETTINGS=/ebook -dNOPAUSE -dQUIET -dBATCH \
   -sOutputFile=small.pdf big.pdf

# /screen = 72 dpi, /ebook = 150 dpi, /printer and /prepress = 300 dpi

If it is still too big

Compression has a floor. A 400-page scanned manual is not going to fit in an email no matter what you do. At that point you have two honest options: split it into parts, or send a link. See reducing a PDF for email for the size limits of every major mail service, and why your PDF is so large if you want to attack the cause instead of the symptom.

Too big even after compressing? Split it into chapters or page ranges instead.

Frequently asked questions

Does compressing a PDF reduce quality?

Only if it downsamples images. Structural compression, font subsetting, object streams, removing unused objects, is completely lossless and changes nothing you can see. Image downsampling does lose detail permanently, which is why you should always keep the original file.

How much can a PDF be compressed?

A text-only PDF usually gives 10-30 percent. A colour scan at 600 DPI can drop 90-95 percent when downsampled to 150 DPI. A PDF full of photographs sits in between, typically 40-70 percent depending on the JPEG quality you accept.

Is it safe to compress a PDF online?

It depends entirely on whether the file is uploaded. Most online compressors send your document to a server. A browser-based compressor runs the same algorithms in your own browser tab, so the file never leaves your computer, the safer choice for anything confidential.

Why did compressing my PDF make it bigger?

Usually because the images were already JPEG-compressed and the tool re-encoded them at a higher quality than the original, or because it added a new revision instead of rewriting the file. Run a lossless structural pass instead, which rewrites cleanly.

Can I compress a PDF without losing the text layer?

Yes. Any tool doing genuine PDF compression keeps text as text. If your file comes back unsearchable, the tool rasterised the pages into images. Test by pressing Ctrl+F and searching for a visible word.