How to Compress a PDF Without Losing Quality

You have a 14 MB PDF that needs to go into a form that only accepts files under 5 MB. Or your scanner produced a 25 MB file for what should have been a two-page letter. Or the PDF is fine but your email provider says "attachment too large" — again.

Compression is the answer, but "compression" is a slippery word. It can mean anything from gently re-compressing streams (barely any size change, zero quality loss) to aggressively rasterizing every page at 40% quality (90% size reduction, most documents look terrible). This guide explains what compression actually does, which levels are safe, and how to hit a specific size target.

Why PDFs get large in the first place

Almost every large PDF has one of three problems:

• Scanned pages. Each page is a full-resolution photo of paper, often 2400×3300 pixels or larger. A 20-page scan can easily hit 30 MB.
• Uncompressed images. Some tools embed photos as PNG or TIFF inside the PDF instead of compressed JPEG. This can inflate a small image to 10× its necessary size.
• Embedded fonts and layers. Rarely the main driver of size, but complex documents with many font subsets and transparency layers can bloat.

The first two are image problems. That's why compression works: images are the compressible part.

The four compression strategies, ranked by safety

1. Metadata removal (zero quality loss)

Stripping author, title, subject, keywords, creation software, embedded thumbnails and unused objects shaves a small amount off the file — typically 1–5%. Quality is completely preserved because nothing visual changes. Safe for every document.

Use Remove PDF Metadata or Optimize PDF (with metadata option enabled).

2. Structure optimization (zero quality loss)

Rewriting the PDF with compressed object streams. Text stays vector, images stay encoded, but the file's internal bookkeeping is more efficient. Reduction is typically 3–15%. Quality preserved. Text stays selectable.

Use Optimize PDF.

3. Image downsampling (lossy but often invisible)

Reduce the pixel density of images embedded in the PDF. A 2400×3300 scan of a page that will be viewed on a screen at 1080×1440 only needs about 1500×2000 pixels. Downsampling to that size reduces the file by 40–70% while remaining visually identical at normal zoom.

Loss depends on how far you downsample. At 150 DPI, most documents look identical to the original. At 100 DPI, text starts to soften on close zoom. At 72 DPI, scanned documents look fine on screen but printed text may show softness.

Use Reduce PDF Image Size or Downsample PDF Images.

4. Page rasterization (lossy, high impact)

Render every page to an image and rebuild the PDF from those images. This is what most "compress PDF" tools actually do. It gives the largest size reduction (60–90%) but has serious trade-offs:

• Text becomes non-selectable and non-searchable.
• Vector diagrams and charts lose crispness.
• Digital signatures are destroyed.
• Links and form fields are removed.

This is what Compress PDF does. For many use cases it is the right choice — email attachments, job applications, quick sharing. But if you need the PDF to remain editable, searchable or signed, it is not.

Choosing the right quality level

OmniPDF's Compress PDF tool offers four levels. Here is what each one does and when to use it:

Light

• Scale 1.5×, JPEG quality 0.88
• Typical reduction: 20–40%
• Visual quality: indistinguishable from the original at normal zoom
• Use for: documents you will archive and occasionally re-open, professional submissions where quality matters

Balanced (default)

• Scale 1.3×, JPEG quality 0.72
• Typical reduction: 40–65%
• Visual quality: virtually indistinguishable for text and simple images; photographs show very slight softening at close zoom
• Use for: most email attachments, job applications, general sharing

Strong

• Scale 1.0×, JPEG quality 0.55
• Typical reduction: 60–75%
• Visual quality: acceptable for on-screen viewing, noticeable softness on printing
• Use for: tight upload limits, mobile viewing, drafts that don't need to look perfect

Extreme

• Scale 0.7×, JPEG quality 0.38
• Typical reduction: 75–90%
• Visual quality: visible softening, text slightly blurry, not suitable for printing
• Use for: last resort when nothing else fits, quick previews

Hitting a specific target size

If you have a fixed limit — say 100KB for a government form, or 200KB for a portal — use the target-size tools:

• Compress PDF to 100KB
• Compress PDF to 200KB
• Compress PDF to 500KB
• Custom Target-Size

These iterate through 7 decreasing scale/quality combinations and stop when the target is met. If the target is not reachable without destroying the file, they return the smallest result and clearly show the actual size in the filename. They never claim to have hit a target they didn't.

100KB is achievable for most text documents up to about 10 pages and for single-page scans. 100KB is not achievable for a 40-page illustrated catalogue — nothing short of severe quality damage will get there, and the tool will tell you so.

When compression cannot help

Some PDFs are already as small as they can be without visible damage. This happens with:

• Text-only PDFs generated by a word processor. Already efficient. Compression may increase size slightly.
• PDFs that were already compressed once. Re-compressing only adds generation loss.
• Heavily optimized vector documents. Diagrams, charts and CAD exports.

In these cases the honest answer is: this file is already small, and the target is not reachable. OmniPDF tells you that instead of returning a broken output.

A practical workflow

1. Check the file size with PDF File Size Checker.
2. Try the lightest pass first: Remove PDF Metadata. Zero quality loss.
3. If still too large, use Compress PDF at Light or Balanced.
4. If the target is still not met, drop to Strong.
5. Only use Extreme as a last resort for drafts and previews.

Always keep the original file. Compression is one-way in the browser — you cannot recover the original quality from the compressed copy.

Summary

Compression quality comes down to the trade-off between size and clarity. Metadata removal and structure optimization are free. Image downsampling costs some invisible quality. Rasterization costs selectable text and every annotation. Pick the lightest option that reaches your target, and never compress a file you still need to edit.