🗜️ Compress PDF

Shrink a PDF to an exact target size, and be told first whether it is a scan (safe) or real text (you lose the text layer). Runs in your browser.

✓ Free✓ No Signup Required✓ Browser-Based

🗜️

Choose a PDF, or drop it here

Compressed in this tab. The document is never uploaded.

What Compress PDF Does

There is no general-purpose way to make a PDF smaller without giving something up. A PDF is already a compressed container — its text is vector drawing instructions, which are extremely compact, and its images are usually JPEG or JBIG2 already. Re-saving it recovers almost nothing. So every tool that reaches a chosen file size does it the same way: by rendering the pages to bitmaps and re-encoding them, which throws away detail and, more importantly, throws away the text.

That is worth stating because the category does not. One of the pages ranking on this term advertises both an exact target size and "lossless quality" at the same time, which cannot both be true. Squeezing a four-megabyte scan into 200 KB discards an enormous amount of information; the question is only whether the information mattered.

And that depends entirely on the document, which is why this tool looks before it acts. pdf.js can read a PDF’s text layer, so the pages are sampled and the extractable characters counted before anything is compressed. A scan has almost none — it is already pictures, so re-compressing costs you nothing you had, and the reduction is dramatic. A born-digital report has thousands, and rendering it to images destroys selection, search, copy and screen-reader access permanently.

The size search then works the way the image compressor’s does: encode, measure the real byte count, adjust. Quality is searched first because re-encoding an already-rendered page is cheap; resolution comes down only when quality alone cannot reach the target. Everything happens in your browser, which for payslips, bank statements and application documents is the point.

How to Use Compress PDF

Add a PDF, or drop it onto the box
Read the notice telling you whether it is a scan or a text document
Choose a target size in KB, or pick a rendering resolution
Turn on greyscale for a black-and-white scan, then compress and save

Formula Used by Compress PDF

What rendering resolution costs in pixels

pixels per page = (widthPt × dpi ÷ 72) × (heightPt × dpi ÷ 72)

widthPt, heightPt: the page size in PDF points, 72 to the inch
dpi: the rendering resolution chosen

Worked example

A US Letter page (612 × 792 pt) at three resolutions.

96 DPI: 816 × 1056 = 0.86 megapixels
150 DPI: 1275 × 1650 = 2.10 megapixels
200 DPI: 1700 × 2200 = 3.74 megapixels

Result: Going from 96 to 200 DPI is 4.3× the pixels, and roughly that multiple in bytes at the same JPEG quality. Resolution is the coarse control; quality is the fine one.

What kind of PDF do you have?

The tool answers this for you by reading the text layer. It decides whether compressing here is harmless or destructive.

Type	How to recognise it	Compressing this way
Scan or photographed document	Almost no selectable text; pages are images already	Safe, and where the big reductions come from
Born-digital text document	Thousands of characters of selectable text	Destructive — the text becomes a picture of text
Mixed	Some pages selectable, some not	The text pages lose their text layer
Already-optimised PDF	Small for its page count, mostly vector	Often comes out LARGER; nothing to win

Choosing a resolution

DPI	US Letter page	Good for
96	816 × 1056	On-screen reading, email, meeting a tight size limit
150	1275 × 1650	The usual balance — readable, zoomable, much smaller
200	1700 × 2200	Documents that will be printed or read closely

How to Read Your Result

If it is a text document, compressing is usually the wrong answer

A PDF of vector text that is too large is nearly always too large because of embedded images or fonts, not because of the text. Removing pages you do not need with the splitter keeps everything selectable and often gets you under the limit on its own. Rasterising a contract to save a few hundred kilobytes trades away search, copy and accessibility for the life of the document.

Greyscale is close to free on a black-and-white scan

JPEG stores brightness and colour separately, and a page of black text on white paper carries no colour information worth the bytes. Discarding it typically takes a useful bite out of the file with no visible change. On anything with photographs or coloured charts, leave it alone.

A larger result is a real answer

When the output is bigger than the input, the file was already efficient and this method has nothing to offer it. That is information, not a failure — it means you can stop looking for a compressor and start looking at what is actually big inside the document.

Limitations & Accuracy Notes

Every page becomes an image. Selectable text, searchability, copy-paste, links and accessibility tagging do not survive. This is inherent to the method, not specific to this implementation.
Around 200 pages is the practical ceiling: each page is rendered to a bitmap and held while the size search runs, and a browser tab runs out of memory beyond that. Split first.
Password-protected documents are refused rather than partially processed.
Bookmarks, form fields, annotations and attachments are not carried into the rebuilt file.
It cannot re-compress the images inside a PDF while leaving the text as text — that needs a full PDF object rewriter, which is well beyond what a browser tab should attempt.
A PDF that is already efficient will get larger, not smaller.

Frequently Asked Questions

Does compressing a PDF lose quality?

Yes, and any tool telling you otherwise is describing something it is not doing. Reaching an arbitrary target size means throwing information away. This works by rendering each page and re-encoding it as a JPEG, so what you lose is image detail and — this is the part that matters — the text layer. Nothing that reaches a chosen byte count is lossless.

Why does it check my document before compressing?

Because the same operation is harmless on one kind of PDF and destructive on another. A scan is already a stack of pictures, so re-compressing costs you nothing you had. A born-digital report is vector text, and rendering it to images destroys selecting, searching, copying and screen-reader access. The tool reads the text layer first and tells you which one you have, because no competing tool on this term makes that distinction.

How do I compress a PDF to exactly 200 KB?

Choose "to a target size" and type 200. The tool renders the pages, then searches the JPEG quality by encoding and measuring until the whole document fits under your number, dropping the resolution only when quality alone cannot get there. It reports the resolution and quality it settled on, so you can see what it cost.

My PDF got bigger. Why?

Because it was already efficient. A text document is vector instructions — extremely compact — and turning those pages into photographs costs far more than it saves. When the result comes out larger, that is the tool telling you the file is already about as small as it usefully gets by this method.

Should I use greyscale?

For a black-and-white scan, yes. JPEG stores colour in separate channels and a page of black text on white paper has no colour information worth keeping, so discarding it is close to free. For anything with photographs or coloured charts, leave it off.

Is my document uploaded?

No. Rendering uses pdf.js and rebuilding uses pdf-lib, both in your browser. The site is a static export with no backend, so there is no server to receive a file — which matters for the payslips, bank statements and application documents people usually need to shrink.

Is there a page limit?

A practical one of 200 pages. Every page has to be rendered to a bitmap and held in memory while the size search runs, and beyond a couple of hundred pages a browser tab runs out of room. Split a longer document first and compress the parts.

References & Further Reading

ISO 32000-1 / PDF 32000 specification (Adobe copy) — Page geometry in points, and how content streams and image XObjects are stored
pdf.js — Mozilla’s PDF renderer — Used both to render the pages and to read the text layer for the scan-or-text check
pdf-lib — Rebuilds the compressed document in the browser
ITU-T T.81 — the JPEG specification — The lossy encoding each rendered page passes through, including the separate chroma channels greyscale discards

By OnlineToolHubs Team • September 2026