Compress PDF
Shrink PDFs without uploading them. Pick a preset, see the estimated output size, preview the quality on a real page from your file before applying.
Compress PDF reduces a PDF's file size by re-encoding embedded images at lower JPEG quality and lower resolution, then repacking object streams and stripping metadata via qpdf-wasm — entirely in your browser. Pick from three presets (Light / Medium / Strong), see live size estimates per preset, and preview the actual quality on a representative page from your file before applying. No upload, no daily limit, up to 100 MB per file. Output is a standard PDF that opens in any reader.
What is Compress PDF?
Compress PDF shrinks the size of a PDF by re-encoding embedded images at lower JPEG quality and lower resolution, then repacking the document's internal object streams and stripping unused metadata. The output is a normal PDF — opens identically in Adobe Acrobat, Preview, browser viewers, anywhere — just smaller. The visual content of every page is preserved; what changes is how aggressively images and structure are packed.
Most online PDF compressors require uploading the entire file to a remote server. For documents that are often confidential — financial statements, signed contracts, medical records, internal manuals — that's a real privacy cost. Pikowl runs the entire compression pipeline locally in your browser using pdf-lib + native canvas re-encoding + qpdf-wasm. The PDF and every byte of its content stays on your device, verifiable in your browser's Network tab.
How does the compression work?
Compression happens in two stages. First, pdf-lib walks the document, locates every embedded image XObject, decodes it via the browser's native canvas API, downsamples it to the preset's pixel cap, and re-encodes it as JPEG at the preset's quality setting. Second, qpdf-wasm takes the rebuilt PDF and runs a final pass that repacks object streams and strips redundant metadata. The combination of both passes is what produces real savings — neither alone does the full job.
Why does it work better on photo-heavy PDFs?
PDF compression value comes mostly from re-encoding embedded images. PDFs that are 80%+ images (photo catalogues, scanned documents, design portfolios, annual reports with charts) shrink dramatically — often to 20–40% of their original size. PDFs that are mostly vector text and diagrams (legal documents, code printouts, ebook exports) shrink only modestly because there's nothing image-heavy to re-encode. The analysis tag above the preset cards tells you which kind of document you have so you set realistic expectations before clicking Compress.
Light, Medium, and Strong — when to use each
Light — Lossless cleanup. Strips metadata (creation tool tags, modification history, custom XMP) and repacks object streams. Image content is untouched — pixel-identical output. Typical savings: 5–15%. Use when you can't tolerate any visible quality change but want the smallest possible spec-compliant PDF.
Medium — Recommended default. Re-encodes embedded images at 75% JPEG quality and downsamples to a 200 DPI cap (high-quality print equivalent). Quality loss is typically not visible on screen at standard zoom. Typical savings: 40–70% on photo-heavy documents, 5–15% on text-only. Use for almost everything.
Strong — Email-friendly. Re-encodes at 50% JPEG and downsamples to a 100 DPI cap (screen-quality only — not suitable for printing). Visible compression artifacts on photographs, but text and vector content remain perfect. Typical savings: 65–85% on photo-heavy documents. Use when getting under an email size limit matters more than image fidelity.
Why is Light called lossless if it changes the file?
We label the Light preset "lossless" because it doesn't touch image content — the visible page content remains pixel-identical to the source. It strips the PDF's mutable metadata (creation tool tags, modification history, custom XMP fields) and repacks the document's object streams, but nothing the reader perceives is changed. Medium and Strong are explicitly lossy on images; Light is not.
The privacy advantage matters more here
Competitor tools (Smallpdf, iLovePDF, PDFCompressor.com, etc.) require you to upload the entire PDF to their servers to compress it. That upload route both leaks the document content (often confidential — financial statements, contracts, medical records) and creates a single point of failure if their service is down or their bucket is breached. Pikowl never sees your file. Compression runs entirely in your browser using pdf-lib + native canvas + qpdf-wasm. You can verify in your browser's Network tab — zero requests to our servers during the operation. The PDF stays on your device.
Why are the size estimates approximate?
The estimated output sizes shown next to each preset are approximations based on the analyzed image-byte ratio of your document and typical compression coefficients. Actual results vary by ±20% depending on document content — vector-heavy PDFs compress less than the estimate; photo-heavy scans often compress more. The ~ prefix is a deliberate signal: don't read these as exact promises. The result state always shows the actual achieved savings.
Limits and behaviour
Maximum file size: 100 MB. Maximum page count: 500 pages. Image count: no hard cap, but documents with more than 200 embedded images may take noticeably longer. Password-protected PDFs are detected and blocked with a link to Unlock PDF. Output is a single compressed PDF; metadata is stripped; pages, layout, fonts, and content remain identical to the input.
Compress PDF: Pikowl vs Smallpdf vs iLovePDF
Compress PDF is the most-used PDF tool in the suite — here's how Pikowl's local-only model compares to the dominant cloud tools.
| Feature | Pikowl | Smallpdf | iLovePDF |
|---|---|---|---|
| Processing location | Local (browser) | Cloud (upload required) | Cloud (upload required) |
| Signup required | No | Optional | Optional |
| Quality preview | Live before/after preview at preset switch | No preview before compression | No preview before compression |
| Max file size (free) | 100 MB | Up to ~5 MB free | Up to ~25 MB free |
| Daily attempt limit | Unlimited | 2 tasks/day on free tier | Free up to 5 tasks/hour |
Last updated April 2026. Smallpdf and iLovePDF figures from their respective free-tier specs at time of writing.
Common Uses
- Email attachment limits: Most providers cap attachments at 25 MB (Gmail), 20 MB (Yahoo), or 10 MB (corporate). Compress photo-heavy reports, signed contracts, and scanned receipts to slip under these limits without splitting the document.
- Mobile data plans: Sharing PDFs over cellular is expensive in many regions; compressing a 30 MB design portfolio to 5 MB before uploading saves real money for the recipient too.
- Long-term archives: Personal document archives (years of bank statements, scanned receipts, medical reports) accumulate fast — compressing each at the Light preset before storage cuts archive size 5–15% with zero visible change.
- Application portals: Loan applications, university submissions, government forms (visa, tax, housing) often impose 5 MB or 10 MB upload caps. Compress before uploading instead of failing the form.
- Mobile-friendly client deliverables: Email a 50 MB design proposal compressed to 8 MB so the client can open it on a phone over a slow connection without downloading for 2 minutes.
- Slack / Discord / Teams sharing: Most messaging platforms have file-size caps (Slack free tier: 1 GB total but suggests <100 MB per file). Compress before sharing to stay friendly to recipients on metered connections.
FAQ
How small can I make my PDF?
It depends on what's in the PDF. Photo-heavy documents (scans, image catalogues, design portfolios) commonly compress 60–85% with the Medium or Strong preset. Vector-only documents (text, diagrams, code listings) compress only 5–15% at any preset — there's nothing image-heavy to re-encode. The analysis tag above the preset cards tells you upfront which kind of document you have.
Will compression affect text quality?
No. Only embedded images are re-encoded. Text remains pixel-perfect at every preset, including Strong, because text is stored as font glyphs, not raster pixels. The same applies to vector diagrams, line art, and any drawing operators in the page's content stream — they're all preserved exactly.
Can I get even more compression?
Strong is our most aggressive preset that still produces visually acceptable images. For extreme compression (presentations on slow connections, archival of low-priority documents), consider exporting individual pages as low-resolution JPEGs first and rebuilding the PDF — but you lose searchable text. For most use cases Strong is the right ceiling.
Is the compressed PDF a different format?
No, the output is still a standard PDF — same .pdf extension, same MIME type, opens in every PDF reader. We use pdf-lib's standard PDF reassembly and qpdf's standard object-stream compression. Both produce strictly spec-compliant PDFs (ISO 32000-2).
What if my PDF is mostly text?
You'll see modest savings (typically 5–15% with the Light preset). Light strips metadata and repacks the document's internal structure — small wins on text PDFs but real ones. The Medium and Strong presets won't help much on text-only docs because there are no images to re-encode.
Why didn't my scan compress as much as expected?
Scanned documents often have already-compressed images (the scanner already JPEG-encoded them). Re-encoding an already-compressed JPEG produces small additional savings. If a scan compresses 20% on Medium, that's expected — the scanner did most of the heavy lifting. Try Strong for additional modest savings.
Is my PDF actually private?
Yes. The entire compression pipeline runs in your browser: pdf-lib reads and rewrites the PDF, native canvas re-encodes images, qpdf-wasm compresses streams. Open your browser's Network tab while compressing — you'll see zero requests to our servers. The only network traffic is the one-time fetch of the wasm runtime, which then caches.