Compress PDF — Keep the Text, Shrink the File
Your words, fonts and layout come through exactly as they are. Only the pictures inside are made smaller. Your document never leaves your device.
Drop your PDF here
or click to select a file
One PDF at a time • Text and fonts are never altered
What is inside this file
How to Compress a PDF
Add your document
Drop it in or browse for it. Nothing is changed yet — the file is only read, and it is read on your own machine.
Read what we found
We report how many pages are scans, how many hold real text, and which carry pictures — because the pictures are where the size is.
Pick a level
Start with balanced. Choose lossless if not one pixel may change, or smallest if the file has to go further than balanced took it.
Download the result
You get the measured before-and-after figures for your own file, and an itemized note of what was removed or re-encoded to get there.
Why Some PDFs Shrink and Others Refuse To
Almost every disappointing result with a PDF compressor comes from the same misunderstanding: people think of a PDF as one kind of thing. It is not. It is a container, and what is inside decides everything about what a compressor can do. Two files of the same page count and the same apparent appearance can behave in completely opposite ways.
Sixty real documents were measured while this tool was being built, and one number explains the whole table below: pictures were 54 percent of every byte across all of them. Everything else — the text, the fonts, the vector drawings, the bookmarks, the form fields — shared the rest. That is why this tool works on the pictures and leaves the rest alone, and why a file with no pictures in it has so little to give.
| What the page holds | Where it came from | What compressing can do |
|---|---|---|
| A photograph of paper | An office scanner, or a phone camera app | A great deal, especially when the scan was made at a needlessly high resolution |
| Real text and vector shapes | Word, Google Docs, LaTeX, an invoicing system | Very little. The content is already stored as compact instructions |
| Photographs placed in a layout | A brochure, a report, a slide deck exported to PDF | Something, but there is a floor — those photographs were compressed once already |
| A scan with an invisible OCR layer | A scanner or archive that ran text recognition | A great deal, and the hidden layer that makes it searchable is kept |
What Actually Gets Made Smaller
Three things happen to your file, and only one of them changes any pixels. The first is hidden bookkeeping: a record of which program wrote the document, when, and often a long editing history nobody has ever read. In one measured file that alone was several megabytes. The second is repetition — the same logo, signature or embedded font stored many times over because each page brought its own copy — which is stored once instead. Both of those leave the pages provably identical, and they are the whole of the lossless setting.
The third is the pictures, and it is the only lossy step. Here the surprise is that lowering quality is usually the weaker lever. A picture inside a PDF has almost always been compressed once already, so re-encoding it at the same dimensions can spend real sharpness for no saving at all — on one measured scan it came out larger at every quality setting worth using. Reducing how many pixels the picture holds is what actually works, and the same picture at 60 percent of its width and height came out 54 percent smaller. That is why the smallest setting changes size and quality together rather than offering a quality slider that flatters and fails.
If none of that reaches a hard ceiling somebody gave you, stop compressing the PDF. Convert the page with our PDF to JPG tool and take the result to the resize in KB tool, which aims at an exact number and can tell you when it got there. If sharpness matters more than size, PDF to PNG keeps every pixel, and image to PDF builds a document back up from images afterwards.
Two Guards Against a Worse File
It cannot hand you something larger
The rewritten file is weighed against your original before you are offered anything. If it is not smaller you get your original back, byte for byte, and the tool says so.
The same check runs per picture
Each picture is kept in its original form unless the re-encoded version is genuinely smaller, so an already tightly packed image is left alone instead of degraded for nothing.
Encrypted files are declined
A password lets us read a PDF but not rewrite it safely. Rather than collect a password and refuse anyway, the tool declines as soon as you drop the file in and tells you to remove the protection first.
Still worth opening the copy
On the two lossy settings, zoom in on a photograph or a signature before you replace anything. Keeping the original until you have looked costs nothing.
Frequently Asked Questions
Is my PDF uploaded anywhere?
No. The file is opened, rewritten and saved entirely inside your browser, and no part of it is sent to a server. This matters more here than on most tools, because the PDFs people compress are usually bank statements, contracts, medical records and scanned identity documents. You can verify it by watching your browser's network panel while you compress, or by going offline once this page has loaded — the tool carries on working.
Will my text still be selectable and searchable?
Yes. Every character, font, link target and page size comes through untouched, because the tool rewrites your file rather than photographing it. Selecting, searching and copying all still work, a screen reader can still read the pages, and an invisible layer left behind by scanning software is kept as well, so a scan that was searchable stays searchable.
How much smaller will my PDF get?
That depends on what your file is made of, and the tool reports the real figure rather than advertising one. Across sixty real documents measured while building this, pictures accounted for 54 percent of all bytes — so files full of photographs and scans give up the most, and a file that is purely text has almost nothing to give. The result you see is measured on your actual file, not projected from a sample.
What is the difference between the three levels?
Lossless changes no pixels at all. It removes hidden bookkeeping data and stores repeated pictures and fonts once instead of many times, so the pages are byte-for-byte identical to look at. The other two re-encode the pictures: balanced keeps them at their current size, and smallest also reduces how many pixels each picture holds, which is the setting that moves a stubborn file.
Why does reducing picture size help more than lowering quality?
Because most pictures inside a PDF have already been compressed once. Measured on a real scanned form, one picture was so tightly packed that re-encoding it at the same size made it larger at every quality worth using. The same picture at 60 percent of its width and height came out 54 percent smaller. Fewer pixels is the lever that works; quality alone often just costs sharpness for nothing.
Why does it say nothing could be saved?
Because some PDFs are genuinely already as small as their contents allow. Text in a PDF is not a picture of letters but a short instruction naming a font, a size and a position, so a twenty-page report can be smaller than one photograph. When that happens the download you get is your original file, byte for byte, rather than a re-saved copy pretending to be an improvement.
Can my file come out bigger than it went in?
No. The rewritten file is measured against your original before anything is offered to you, and if it is not smaller the tool hands back the original untouched and says so. That check exists because a compressor that returns a larger file while reporting success is worse than one that admits defeat. The same guard runs on every individual picture inside the file.
Can I compress to an exact size like 100 KB or 500 KB?
No, and no honest browser tool can promise that for a PDF. Every file has a floor set by its own contents, and a document already full of compressed photographs can sit several megabytes above any target you name. Choose the smallest setting to go as far as this approach goes, read the real figure, and if you need to hit a hard ceiling convert the page to an image instead.
Does it work with password-protected PDFs?
A password is enough to read a PDF but not to rewrite one safely, so encrypted files are declined the moment you drop one in, before you are asked for anything. Inside an encrypted PDF every stream is scrambled, and rebuilding it without handling that scrambling produces a file that opens to blank pages. Remove the protection in whatever application applied it, then compress the unprotected copy here.
More Tools
Last updated: