Skip to main content

Engineering Lossless PDF Compression: Vector Preservation vs. Raster Downsampling

How we built a zero signup PDF compression pipeline that cuts document weight by up to 80% while keeping crisp typography, ATS compatibility and clickable links.

Updated 3 min readMarkdown

Every developer and remote worker eventually runs into annoying document size caps. Maybe a 15 MB PDF portfolio gets rejected by a 5 MB applicant tracking system (ATS), a 30 MB slide deck gets blocked by Gmail's 25 MB MIME limit, or a scanned contract times out on a government portal.

Most common workarounds, like printing to a virtual PDF printer or running basic browser canvas flattening, wreck the internal structure of the file. They turn sharp vector font glyphs into blurry bitmaps and strip out interactive /Annots link dictionaries.

At UtilifyAI, we built our Free Online PDF Compressor and wrote our 2026 Master Guide to Lossless PDF Compression around one simple rule: never touch vector geometry; only optimize embedded bitmaps and object streams.


Understanding the Internal Object Graph of a Bloated PDF

Inside a .pdf container, page content is split into separate object streams:

  1. Vector and Typography Streams: Stored as mathematical Bezier curves and character codes. These take up just a few kilobytes and scale to any resolution without blurring.
  2. Embedded Font Programs (/FontFile2, /FontFile3): Often unsubsetted, carrying thousands of unused Unicode glyphs that add 2 to 4 MB of dead weight.
  3. Raster Image Streams (/XObject /Subtype /Image): Raw 300 to 600 DPI pixel arrays dropped in by Canva, Word, Keynote, or mobile scanners. These account for 75% to 90% of total PDF weight.
  4. Cross-Reference (xref) and Metadata Bloat: Incremental save histories, XMP thumbnails and app-specific PieceInfo dictionaries.

Our 3-Tier DPI Optimization Architecture

Instead of forcing every file through a single aggressive filter, the UtilifyAI PDF Compressor runs documents through a two-stage Ghostscript (pdfwrite) + pikepdf pipeline with three clear presets:

Preset Mode Target Image DPI Typical Size Reduction Best Use Case
Recommended 144 DPI (/ebook) 40% to 70% smaller Resumes, slide decks, email attachments, Retina displays
Maximum 72 DPI (/screen) 70% to 90% smaller Strict 1MB to 2MB government, university & ATS portals
Basic 300 DPI (/printer) 20% to 40% smaller High-resolution commercial printing & archival handoffs

During processing, the pipeline guarantees three things:

  • Zero Text Rasterization: Vector fonts and SVG paths stay 100% selectable, searchable and ATS-parseable.
  • Font Table Subsetting: Unused glyphs get pruned while keeping the exact visual look of the document.
  • Interactive Link Preservation: Hyperlinks, bookmarks and document outlines stay 100% clickable.

Try the Live Tool & Read the Full Breakdown

Whether you need to shrink a PDF under 2 MB right now or want to see our benchmark tables across Gmail, Outlook, Workday and Canvas upload limits:

About the author

No comments yet. Be the first to comment.