The Complete Guide to PDF Compression
Two engines, honest expectations, and the privacy model — everything you need to shrink a PDF in your browser.
1. Why PDFs get so large
PDFs are often far larger than the visible content they contain warrants. A 40-page contract pack can balloon past 20 MB because of redundant font embedding, uncompressed streams, and legacy object tables. PDF compression is the process of rewriting a PDF so the same document occupies less space.
2. Two compression engines
This tool offers two engines, and which one is right depends entirely on what the source PDF contains:
- Structural rewrites the PDF’s object table and bundles indirect objects into compressed object streams. It is lossless — text, fonts, images, and annotations are byte-identical in the output. It works best on text-heavy PDFs (contracts, reports, ebooks) which typically shrink 15–35%. It does almost nothing on scanned PDFs, whose size is dominated by embedded raster images.
- Raster renders each page to a JPEG at your chosen DPI, optionally grayscale, and rebuilds the PDF from those JPEGs. It is the only way to actually shrink scanned / image-heavy PDFs. Text becomes non-selectable in the output — that is the trade for the size win. The Aggressive preset at 80 DPI grayscale routinely shrinks a 20 MB scan to under 2 MB.
3. What you can realistically expect
- Text-heavy PDFs (Structural) — 15–35% shrink, lossless.
- Scanned PDFs (Raster, Balanced) — 60–80% shrink, near-identical to the eye.
- Scanned PDFs (Raster, Aggressive + grayscale) — 80–95% shrink, suitable for archiving or email.
- Already-optimised Office/chrome-PDFs (Structural) — may grow slightly; the tool reports this honestly.
4. Is this private?
Completely. Both engines run inside your browser tab usingpdf-lib and pdfjs-dist. The file is loaded into your device’s memory, processed, and downloaded straight back — nothing is transmitted over the network. For NDAs, financial statements, and medical records, client-side processing is the only acceptable option.
Match the Engine to the Source
Pick Structural for digitally-authored text PDFs where fidelity matters, and Raster for scanned or image-heavy documents where file size is the problem. Either way the work happens entirely in your browser, the result is honestly reported, and your file never leaves the device. Drop a PDF above, choose the engine that matches the source, and download a leaner document in seconds.