ConvertBox Logo
ConvertBox

Why PDF files are so large, and which fix actually works

A PDF is large because of its images, its fonts or data it no longer needs, and each cause has a different remedy. Structural compression removes dead weight only; shrinking a scan means lowering the resolution of its images.

ConvertBoxPublished

What a PDF is made of, and what each part weighs

A PDF is a collection of numbered objects: one content stream per page holding drawing instructions, plus the fonts and images those instructions refer to. The instructions themselves are tiny. A full page of typed text is a few kilobytes once compressed, because the file stores character codes and positions, not a picture of the letters.

Size comes from three other places: images, embedded fonts, and data the file carries but does not need. Knowing which one dominates your file decides the fix, because a method that works on one cause does nothing for another.

Scanned pages: the pixel arithmetic behind big files

A scanner does not produce text; it produces one photograph per page. An A4 sheet is 8.27 x 11.69 inches, so at 300 DPI it becomes 2480 x 3508 pixels, about 8.7 million. Stored as 24-bit colour that is 26 MB of raw data per page before compression. JPEG compression brings it down a long way, yet a colour page still typically lands between a few hundred kilobytes and more than a megabyte.

Halving the resolution quarters the pixel count, and dropping colour cuts the raw data to a third. Pure black-and-white scanning, where each pixel is one bit, is smaller still and pairs with compression schemes designed for text pages, which is why a fax-style scan of a text document can be a few tens of kilobytes per page. For reading on a screen, 150 DPI is comfortable; 300 DPI matters when the pages will be printed again or run through OCR.

Scan settingPixels per A4 pageRaw data before compression
300 DPI, colour2480 x 350826.1 MB
300 DPI, greyscale2480 x 35088.7 MB
150 DPI, colour1240 x 17546.5 MB
300 DPI, black and white (1 bit)2480 x 35081.1 MB

Embedded fonts, repeated resources and leftover data

To look the same everywhere, a PDF usually embeds the fonts it uses. Good exporters embed a subset, only the glyphs that appear in the document, which costs tens of kilobytes per font. Embedding whole fonts is where it gets expensive: a Latin font is often a few hundred kilobytes, while a Chinese, Japanese or Korean font with tens of thousands of glyphs can run to several megabytes or more. A two-page memo with no pictures that weighs several megabytes is nearly always carrying a fully embedded font of that kind.

The second source is repetition. A well-built PDF stores a logo once and lets every page point at it; a careless generator embeds a fresh copy on each page, so a 150 KB letterhead on 100 pages adds 15 MB. The third is history. PDF allows incremental saving, where an editor appends changed objects to the end of the file and leaves the old versions in place. After many rounds of annotating and re-saving, a file can contain several generations of pages nobody will see.

How to tell which cause you are dealing with

  • Divide the file size by the page count. Text pages average tens of kilobytes; anything near or above 500 KB per page points to images.
  • Try to select a word. If the cursor draws a box over the whole page instead of highlighting text, the page is a scan.
  • Open the document properties and look at the font list. Entries marked as embedded subset are cheap; a long list of fully embedded fonts is not.
  • Check the history. A file that has been annotated, signed and re-saved many times, or that came out of an old system, may be carrying leftover objects.
  • Compare with the source. If the word-processor or slide file is 1 MB and its PDF is 30 MB, the export settings kept every picture at full resolution.

Which fix addresses which cause

Re-exporting from the original document is the strongest fix whenever the source exists, because the exporter can make every decision again: image resolution, JPEG quality, font subsetting. Presets named for print or press keep images at 300 DPI or leave them untouched; presets named for screen, web or minimum size downsample them to roughly 100 to 150 DPI.

CauseFix that worksWhere it is done
Pictures exported at full resolutionRe-export from the source file with a smaller-size or on-screen presetThe export dialog of the authoring application
Scans at high DPI or in colourRescan at 150 DPI or in greyscale, or downsample the images inside the PDFThe scanner settings; desktop software such as Ghostscript or Acrobat
Fully embedded fontsRe-export with font subsetting enabledThe authoring application or a desktop PDF optimiser
Leftover objects from repeated savesRebuild the file structureA structural compressor, including browser-based ones
Pages the recipient does not needDelete them, or split the documentPage-level tools; no quality is lost

What structural compression does, and what it cannot do

Two different operations share the name PDF compression. Structural compression rewrites the file without dead weight and changes nothing you can see. Lossy compression downsamples and re-encodes the images, trading sharpness for size. ConvertBox's Compress PDF tool performs only the first: it copies every page into a new document, drops data that is no longer referenced, and saves with compact object streams. It does not downsample or recompress images, and it has no quality levels.

The consequence is predictable. A file edited and re-saved many times, or exported by older software, gets smaller. A scan barely changes, because nearly all of its bytes are image data that is copied untouched; when the rebuilt file is not smaller the tool reports 0% and the original is the one to keep. A rebuild of this kind also leaves out bookmarks, fillable form definitions and attachments, and invalidates digital signatures, so check the output before replacing a form or a long manual. To shrink the images inside a PDF while keeping selectable text, use desktop software that downsamples them, such as Ghostscript or Acrobat.

When you cannot re-export: pages, parts and the image route

Removing what the recipient does not need is lossless and often enough. Deleting appendix pages in a page organiser, or splitting a document into parts that each fit under the limit, leaves the remaining pages exactly as they were. One caveat applies to text documents: a font shared by all pages is stored once in the original but copied into every part, so the parts can add up to more than the whole.

For a scan with no better source, the remaining route is through images: export the pages as JPEG, compress them, and build a new PDF from the results. ConvertBox's PDF to JPG tool renders pages at 144 DPI, about half the linear resolution of a 300 DPI scan. The price is that any selectable text becomes pixels, so this suits scans, not text documents. Finally, leave a margin under email limits: attachments are encoded for transport in a way that adds about a third to their size, so a 20 MB file can be rejected by a 25 MB limit.

Tools mentioned in this guide

Related guides

All guides