Most PDF tools that call themselves optimizers are really just image compressors. They re-encode your JPEGs at lower quality, report a smaller file size, and call it done. That approach works — but it misses the structural bloat that plagues many PDFs. A slide deck that embeds the same 200 KB logo on every slide carries that image 40 times as 40 separate objects. Compressing each copy saves proportionally less than simply storing the image once and referencing it forty times. That's deduplication — and most "optimize" tools don't do it.
Linearization is the other commonly missed technique. A non-linearized PDF stores pages in the order they were created or edited — meaning a viewer must download the entire file before displaying any of it. Linearization reorders the byte layout so page 1 data leads the file. The difference is perceptible: a linearized 20 MB catalog shows its first page in under a second on a slow connection; the same file without linearization shows nothing for several seconds. For PDFs served on websites, in email, or via CDN, linearization is the single most impactful structural change you can make.
Structural optimization solves different problems than compression. Here are the six use cases where linearization, deduplication, and font subsetting make the most tangible difference.
PDFs embedded in web pages via PDF.js or native browser viewers load dramatically faster when linearized. The first page renders before the full file downloads, keeping users engaged instead of staring at a loading indicator.
Gmail caps attachments at 25 MB; many corporate mail servers are stricter. A branded report bloated with repeated logo images often shrinks significantly through deduplication alone — without any visible quality change to the recipient.
Quarterly reports, board decks, and client proposals typically contain repeated brand assets and embedded font sets far larger than the document actually needs. Optimization reduces both without altering a single word or pixel the reader sees.
Digital publications (ebooks, catalogs, lookbooks) distributed via download links or embedded viewers reach readers faster and feel more responsive when linearized. First-chapter display while the rest downloads is the difference between a bounce and a read.
Smaller, linearized PDFs are noticeably faster to open on mobile devices, especially on cellular connections. Font subsetting and deduplication together often cut file size enough to move a PDF from 'slow to open' to 'opens immediately' on a mobile browser.
Document archives accumulate PDFs edited many times, each edit round leaving orphaned objects and duplicate resources. Optimization removes this accumulated debt, producing clean, compact files that are easier to store, back up, and migrate across systems.
Structural optimization in a browser — without a server — requires solving non-trivial engineering problems. Here's what that effort means for you.
Linearization: LuraPDF uses pdf-lib to write a new PDF byte stream with the first page's objects at the beginning of the file, followed by a linearization dictionary that gives progressive-download readers a map to the rest of the content. This doesn't change what's in the PDF — it changes the order in which bytes are laid out on disk, enabling partial rendering before full download completes. Deduplication: The engine computes a content hash of every embedded image XObject. Objects with matching hashes are consolidated — the first instance is kept, and every subsequent reference in the page content stream is rewritten to point to the same shared object. The savings are proportional to how many times a given image was embedded separately.
Font subsetting: For each embedded font, LuraPDF analyzes which Unicode code points are actually referenced in the document's text streams. It then rebuilds the font's glyph table to contain only those code points, discarding the rest of the character set. For CJK (Chinese, Japanese, Korean) fonts that embed thousands of glyphs for documents using only a few hundred, the size reduction is dramatic. Unused object stripping: After deduplication and subsetting, the engine walks the PDF's cross-reference table and marks every object reachable from the document catalog. Unreachable objects — deleted pages, removed form fields, old revision snapshots — are excluded from the new file's xref, effectively removing them from the output.
| Feature | LuraPDF | Smallpdf / Sejda | Adobe Acrobat |
|---|---|---|---|
| Linearization (fast web view) | Yes — full linearization | Partial — some tools only | Yes |
| Image object deduplication | Yes | Rarely included | Yes |
| Font subsetting | Yes | Rarely included | Yes |
| Processes file in browser (no upload) | Yes — fully local | No — server upload required | Yes — installed software |
Optimization works best when you understand which technique targets which problem. Match the approach to your file type for the best results.
Linearize if serving PDFs on a website or via email link — first-page rendering speed is the most user-visible benefit for web-hosted documents
Run deduplication first on slide decks and branded reports — these files almost always contain repeated logo and background images that deduplication removes without any quality tradeoff
Enable font subsetting when distributing PDFs externally — especially documents that used design fonts with large character sets, or anything created in InDesign or Illustrator
Pair optimization with Compress PDF for maximum size reduction — optimize removes structural bloat first, then compress re-encodes images at lower quality for an additional reduction
Always keep an unoptimized backup if the PDF is a shared template — subsetting removes unused glyphs, which means adding new text in those characters later requires re-embedding the full font
Test the optimized PDF in your target viewer after — linearized PDFs are standard, but some enterprise document management systems have their own PDF handling quirks worth verifying
Drop your PDF and run all three structural optimization techniques in your browser. Faster web loading, smaller email attachments, cleaner files — no upload, no account, no watermark.