Guide 8 min read Updated 2026

How to compress a PDF without losing quality

Most “PDF compressor” pages hide the only decision that matters behind a single button. This guide explains that decision, so you can keep a contract readable and still get under an email limit.

A PDF is a bag of parts: text, fonts, vector drawings, images, annotations and metadata, all tied together by a small index. “Compressing” it means shrinking one or more of those parts — and each part behaves differently. That is why one PDF goes from 18 MB to 1.4 MB while another barely moves.

Step 1: Find out what is making the file big

Before touching a preset, look at the source:

  • A scanned document is essentially one giant photo per page. Expect huge savings from image compression — often 80–95%.
  • A text document exported from Word, LaTeX or InDesign is mostly fonts and vector data. Rasterising it will usually make it bigger, not smaller.
  • A presentation export is mixed: sharp vector text plus full-resolution images. Image compression helps; text usually survives if the tool keeps the text layer.
  • A PDF with embedded high-DPI photos (catalogues, brochures) is the best case. Photos at 600 DPI contain four times the data of the same photo at 300 DPI for no visible benefit on screen.

A quick signal: if you zoom in to 400% and the text stays crisp with no pixel edges, the page is vector-based. If it blurs, the page is a scan.

Step 2: Pick your trade-off

MethodWhat happensTypical savingText stays selectable?
Lossless rebuildMetadata removed, structure re-packed, unused objects dropped2–15%Yes
Image re-encoding (150 DPI)Pages rendered and saved as JPEG at ~70% quality60–95% on scansNo
Aggressive (96 DPI)Same, with fewer pixels and lower quality85–97% on scansNo
Grayscale + re-encodeColour removed before JPEGadds 10–40%No

The honest summary: you cannot get a 90% reduction while keeping a searchable text layer in a browser-only workflow. Anything that claims otherwise is either re-encoding images underneath (and quietly dropping text) or measuring a resized file.

Step 3: Match the preset to the job

Contracts, invoices, anything you may need to search

Use the lossless rebuild. It removes author metadata, clears the subject and keyword fields, re-packs the object structure and keeps every glyph exactly where it was. A 3 MB contract might only drop to 2.7 MB — but it stays searchable, copyable, screen-reader friendly and legally identical in appearance.

Scanned paperwork you only need to read

150 DPI at around 70% JPEG quality is the sweet spot. At that setting, A4 text is still clearly legible on a phone and a 12 MB scan typically lands between 600 KB and 2 MB. Going below 120 DPI is where fine print starts to fall apart.

Hitting a hard attachment limit

Most mail servers cap attachments at 10 MB or 25 MB. If 150 DPI is not enough, drop to 96 DPI and enable grayscale — that combination is what makes a 20 MB scan fit into a 1 MB limit. Test the result on a phone before sending a legal document this way.

Step 4: Check the output, always

Three things to verify in the compressed file:

  1. Page count is unchanged.
  2. Page size (A4, Letter) is unchanged — printing should not shift the margins.
  3. Try selecting text. If the selection works, you used a lossless method; if not, the page was rasterised and the text layer is gone.

What about “optimise for web” exports?

Many desktop tools offer their own compression. They work well, but they usually optimise the images inside the PDF rather than re-rendering pages — which is the better approach when you need both a text layer and a small file. A browser-only tool cannot easily do that yet, because re-encoding individual embedded images means parsing the PDF’s internal object graph and replacing streams in place. If that matters to you, use a desktop tool like Ghostscript:

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.7 \
   -dPDFSETTINGS=/ebook -dNOPAUSE -dBATCH \
   -sOutputFile=small.pdf input.pdf

The /ebook preset targets 150 DPI images while preserving text — the same idea as the Recommended preset here, but applied to the original image streams.

Doing it without uploading the file

PDFs are often the most sensitive document you own: bank statements, medical reports, signed contracts. The MySelfPDF PDF Compressor runs the renderer and the PDF writer inside your browser, so the file never leaves your device. It also tells you when a rebuild would make the file bigger, instead of quietly handing you a worse result.

Try it on your own PDF

Four presets, grayscale option, before/after report. Nothing is uploaded.

Open PDF Compressor