PDF Compressor

Compress PDF files online for free. Reduce PDF size without losing quality. No sign-up required.

✓ Free✓ No sign-up✓ Works in browser

Drop a PDF here or click to browse

Compresses metadata and optimizes object streams

Adobe Acrobat

The world standard for PDFs

Partner

Edit, sign, compress, merge, protect PDFs. 500M+ users. 7-day free trial.

Start Free Trial

How to Use This Tool

Upload Your PDF

Click 'Upload PDF' or drag and drop your file. Supports PDFs up to 100 MB.

Select Compression Level

Choose Strong (maximum reduction), Recommended (balanced quality), or Low (high quality, modest reduction).

Download Compressed PDF

Click Compress PDF. The before/after size comparison is shown. Download your smaller PDF instantly.

Related Tools

PDF Merger

PDF Splitter

PDF to Word Converter

PDF Password Remover

PDF Editor

Frequently Asked Questions

How much can a PDF be compressed?

Results vary by content. PDFs with many images can shrink 70–90%. Text-heavy PDFs typically compress 10–25%. The tool shows exact savings after compression.

Does compression reduce PDF quality?

Strong compression slightly reduces embedded image resolution (from ~300 DPI to ~150 DPI). For screen and email use, the difference is barely noticeable. Text quality is never affected.

Is the compressed PDF still a valid PDF?

Yes. The output is a fully valid PDF file that can be opened in any PDF reader, printed, and further edited.

Can I compress a password-protected PDF?

No. Remove the password protection first using our PDF Password Remover, then compress the file.

Is this PDF compressor free?

Completely free with no daily limits, no sign-up, and no watermarks. All files are automatically deleted within 1 hour.

About PDF Compressor

Your proposal is 34MB because the designer dropped four full-resolution iPhone photos at 4032x3024 into the appendix, and the client's procurement portal hard-caps uploads at 10MB. Or you need to email a 60-page scanned contract but Outlook rejects anything over 20MB and corporate MTAs downstream often drop anything over 10MB. This compressor re-encodes embedded images through a canvas downsample (to 150 DPI for screen reading, 72 DPI for web preview, or a custom target), subsets embedded fonts so you are not carrying the full CJK glyph table to render three Latin letters, and rewrites object streams with deflate compression. Everything runs inside the browser — pdf-lib.js parses the document, a Canvas 2D context handles JPEG re-encoding at your chosen quality, and the rewritten bytes stream out as a Blob download. Typical compression ratios are 40 to 75 percent for image-heavy documents and 5 to 15 percent for text-only PDFs that are already well-structured. Works on files up to roughly 200MB before browser memory pressure becomes the limiting factor.

When to use this tool

Hitting a procurement portal's upload cap

A sales engineer's 34MB proposal with embedded product photos will not upload to Ariba's 10MB-per-attachment limit. Compress at 150 DPI JPEG quality 70 and the file drops to 7.8MB with no visible quality loss in the screen-viewed version the buyer will actually read.

Emailing a multi-page scanned contract

A 60-page contract scanned at 300 DPI from a Canon MFP produces a 48MB PDF. Gmail's 25MB attachment cap rejects it. Downsample scanned images to 150 DPI and the file drops to 9MB, comfortably under the cap while staying legible on any screen the recipient will use.

Preparing a portfolio for web upload

A designer has a 28MB PDF portfolio with embedded 8-megapixel renders. LinkedIn, Dribbble, and most job-board portals limit attachments to 10MB. Compress aggressively at 72 DPI since the file will only be viewed on-screen, dropping to 4MB while preserving color fidelity.

Archiving client files under storage quotas

A solo accountant with Google Drive's 15GB free tier receives 400 client returns per season averaging 12MB each. Batch-compress to 150 DPI, cutting the total archive from 4.8GB to around 1.8GB, extending the free tier through two more seasons before needing a paid plan.

Shrinking a PDF for fax or MFP upload

A legal assistant has to upload a 40-page document to an old Ricoh MFP that chokes on files over 15MB. Compress to 150 DPI black-and-white (grayscale scans) and the file drops to under 8MB, within the MFP's buffer, and faxes through without a re-scan.

How it works

1
Images are extracted and re-encoded via canvas
pdf-lib walks the page content streams looking for XObject image references. Each image is decoded into a Canvas 2D context, scaled to your target DPI (we compute the target pixel dimensions from the page's physical size in points), then re-encoded as JPEG at the chosen quality. The original image object is replaced in the PDF with the smaller re-encoded version, and the cross-reference table is updated.
2
Fonts are subsetted to only used glyphs
A full embedded Calibri with the CJK extension can be 6MB. When you only use 80 unique characters, we walk the content streams to find the actual glyph set in use, then rewrite the font object with a subset that only carries those glyphs plus the /ToUnicode CMap. Subsetting typically reclaims 60 to 90 percent of font overhead in documents that embed large Unicode fonts for a short run of Latin text.
3
Streams are re-deflated with optimal settings
Many PDFs in the wild use suboptimal deflate settings — some are uncompressed because the producing tool never compressed them, others use level 1 for speed. We re-encode every stream with level 9 deflate, which typically gains another 5 to 15 percent on text-heavy documents. Object streams (PDF 1.5+) are also re-packed to bundle small indirect objects together, further reducing per-object overhead.

Honest limitations

· Vector-only PDFs (contracts with no images) will compress 5 to 15 percent at most because there is little to reclaim beyond font subsetting and stream re-deflation.
· Scanned documents without an OCR layer stay as images — the compressor shrinks them but cannot convert image-of-text into selectable text; for that, OCR is needed first and OCR requires server-side processing in most tool chains.
· Aggressive JPEG quality below 50 produces visible blocking artifacts in images with fine detail (logos, diagrams with thin lines, faces) — preview the output before committing if quality matters for the audience.

Pro tips

Pick DPI based on consumption, not source

A 300 DPI scan is appropriate for archival or print, but wasteful if the document will only ever be read on a 1920x1080 monitor at 96 DPI effective. Rule of thumb: 300 DPI for anything that will be printed or submitted to a court clerk who prints, 150 DPI for screen-reading including tablets and phones, 72 DPI for web preview thumbnails where file size dominates. Downsampling from 300 to 150 DPI cuts image payload by 75 percent (quadratic in linear DPI) — always audit the consumption target before picking a quality level.

Use grayscale conversion on scanned text pages

Scanned contracts and forms are almost always greyscale content (black ink on white paper) stored as RGB JPEGs at three times the necessary size. Converting to grayscale during re-encoding cuts the bytes per pixel from 3 to 1 at identical visible quality. The compressor does this automatically when every pixel in an image falls within 5 percent of R=G=B thresholds, but you can force it for borderline color scans (faint yellowing on old paper) where the color channel carries no useful information.

Run the pdf-splitter first for extraction tasks

If you only need 20 pages of a 500-page source, split those 20 out before compressing. Compressing the whole 500MB source and then splitting would burn three minutes of CPU on pages you will discard. Splitting first reduces the working set to what you actually need, lets compression complete in 10 seconds instead of three minutes, and avoids the edge case where browser memory pressure crashes the tab before compression finishes. This matters more than it sounds — low-RAM devices (8GB Chromebooks, older iPads) will hit heap limits on 500MB inputs.

Frequently asked questions

How much compression should I expect on my file?

Depends almost entirely on image content. A 30-page marketing PDF stuffed with 4000x3000 product photos can compress 70 to 85 percent by downsampling to 150 DPI because the source is hugely oversized for screen reading. A 30-page legal contract with one signature image and otherwise pure vector text will compress 8 to 15 percent — there is little to reclaim. Look at the PDF in Adobe Acrobat's File → Properties → Fonts and Images to get a sense of the image payload before compressing; if images dominate, expect big wins.

Will compression make my PDF look worse?

At 150 DPI and JPEG quality 70-80, most readers cannot distinguish compressed from original on a screen. At 72 DPI (thumbnail-class) you will see softening on zoom and diagonal line aliasing in diagrams. At quality below 50, JPEG blocking becomes visible on solid-color regions. The compressor shows a before/after size summary so you can verify the gain is worth the quality trade, but always preview the output in Adobe Reader at 100 percent zoom before sending to a client or court. For archival copies, keep the uncompressed original in parallel.

Does compression work on scanned documents?

Yes, and scans are where compression wins biggest. A 60-page scan at 300 DPI averages 50MB because each page is a ~800KB image. Downsampling to 150 DPI and converting to grayscale (most scanned contracts are effectively black-and-white) typically cuts to under 10MB. What compression cannot do is add a text layer to a scan — that requires OCR, which is a separate step that typically needs server-side processing (Tesseract running in the cloud, Adobe's cloud OCR, or a WASM OCR that requires loading a ~30MB model). Compress first, OCR separately if searchability is needed.

Is my PDF uploaded to a server for compression?

No. Compression runs in your browser tab using pdf-lib.js and a Canvas 2D context for image re-encoding. The file you select is read via FileReader into an ArrayBuffer, processed entirely in JavaScript heap memory, and the compressed output is delivered as a downloadable Blob. No network request carries the file contents. This matters when the PDF contains client financials, health information subject to HIPAA, or legal documents under attorney-client privilege where server-side SaaS tools would create a compliance exposure you cannot resolve after the fact.

Why did compression barely shrink my text-only PDF?

Text-only PDFs are already efficient. Modern PDF producers (Word, Google Docs, LaTeX, Adobe InDesign) compress content streams with deflate, subset embedded fonts to only used glyphs, and use object streams to reduce per-object overhead. There is little left to reclaim — maybe 5 to 15 percent from re-deflating with level 9 and re-packing object streams more tightly. Big compression wins come from image downsampling and font subsetting, and pure-text PDFs have nothing to downsample. If your text PDF is unexpectedly large (50+ MB for 100 pages), the producer likely embedded full CJK fonts unnecessarily — subsetting will help there.

Compression is usually a late step in a PDF workflow. After the pdf-merger combines multiple documents into one, the result often exceeds the 10MB email or portal cap and goes through the compressor to fit. When a single page needs rotation correction, rotate-pdf handles it before compression so the rotated images re-encode only once. For scanned image stacks, images-to-pdf assembles them first at the target DPI, which can skip the compression step entirely if you choose output DPI sensibly up front. And if the compressed output needs password protection before sending to a client, protect-pdf adds the user password as the final step.