📋 Contents
The Three Types of PDF Compression
When you click "Compress," these three things happen (or should happen):
1. Image Downsampling and Recompression
This is where 90% of file size reduction comes from. PDFs contain embedded images — photos, logos, screenshots. A 24MP photo embedded in a PDF at 300 DPI might be 15MB by itself. The compressor:
- Reduces the resolution (DPI) of each image. A 1200 DPI photo becomes 150 DPI — visually identical on screen, 1/64th the pixels
- Re-encodes images with more aggressive compression. JPEG quality 85 → 60, or PNG → JPEG conversion for photos that don't need lossless encoding
- Strips color profiles from images that don't need them (sRGB is fine for screen viewing; AdobeRGB profiles add KB to every image)
2. Object Stream Compression
PDFs store content as numbered "objects" — text blocks, font descriptors, page descriptions, cross-reference tables. A typical 50-page PDF has thousands of objects. The compressor:
- Combines small objects into "object streams" (compressed containers)
- Removes duplicate objects — two pages using the same background image? Keep one copy, reference it twice
- Compresses uncompressed content streams using FlateDecode (zlib). Some PDF creators leave content streams uncompressed for compatibility
3. Structural Cleanup
This removes data that exists in the file but isn't needed for display:
- Unused pages, objects, and fonts — PDF editors often leave deleted content in the file as orphaned objects
- Metadata bloat — some PDFs have 50KB of redundant XMP metadata from multiple editing applications
- Embedded thumbnails — some PDF creators embed preview images. They're redundant and can add 50-100KB per page
- JavaScript, form data, comments — if you're archiving or distributing a final document, you don't need interactive elements
What Actually Works Per Document Type
| Document Type | Typical Size | Best Approach | Expected Reduction |
|---|---|---|---|
| Photo-heavy (portfolio, catalog) | 20-80 MB | Aggressive image downsampling: 300→150 DPI, JPEG quality 85→60 | 70-90% |
| Text-heavy (report, ebook, contract) | 1-5 MB | Object stream compression + font subsetting. Images are minimal — don't downsample aggressively | 30-50% |
| Scanned document (OCR'd) | 10-50 MB | Convert to黑白 (black & white) if color isn't needed. JBIG2 compression for text regions, JPEG for photo regions (MRC compression) | 60-85% |
| Presentation (PPT export) | 5-15 MB | Downsample slide backgrounds, strip unused master slides, remove speaker notes from PDF export | 40-60% |
| CAD/Engineering drawing | 2-20 MB | Vector optimization: merge identical paths, simplify bezier curves with excessive control points | 20-40% |
Exact Settings for Best Results
For general-purpose compression (most documents), these settings produce good results:
| Setting | Conservative (visible quality) | Aggressive (smallest file) |
|---|---|---|
| Color image DPI | 150 DPI, JPEG quality 85 | 100 DPI, JPEG quality 50 |
| Grayscale image DPI | 150 DPI, JPEG quality 75 | 100 DPI, JPEG quality 50 |
| Monochrome image DPI | 300 DPI, CCITT G4 | 200 DPI, JBIG2 lossy |
| Font strategy | Subset (embed used chars only) | Subset + convert Type3 to Type1 |
| Cleanup | Remove unused objects, metadata | Remove all: metadata, thumbnails, JS, forms, comments, private data |
💡 Rule of thumb: For on-screen reading (email attachments, web download), 150 DPI images look identical to 300 DPI. The screen can't display higher resolution than its pixel grid. Only preserve 300+ DPI if the PDF will be printed.
Common Mistakes That Make PDFs Bigger
- Embedding full fonts instead of subsets. A full OpenType font can be 2-5MB. If your document only uses the letters A-Z, a-z, 0-9, and basic punctuation, subsetting reduces the embedded font to ~15KB. Always subset fonts unless the recipient needs to edit the text
- "Print to PDF" instead of "Save as PDF." Print-to-PDF drivers often embed images at 1200 DPI with no compression. Same document, "Save as PDF" might be 2MB vs "Print to PDF" at 45MB
- Embedding the same image multiple times. If you copy-paste the same logo onto 50 pages, some editors embed it 50 times instead of referencing one copy. Deduplication fixes this
- Using PNG for photos. PNG is lossless. A photo saved as PNG in a PDF is 5-10× larger than the same photo as JPEG. Use PNG for screenshots and logos with sharp edges; use JPEG for photographs
Compress your PDF now:
Compress PDF →