Why PDF files get large, and how to shrink them

Two PDFs can both be ten pages long and differ in size by a factor of a hundred. One is a 150 KB report typed in a word processor; the other is a 15 MB scan of the same report run through a photocopier. Nothing is wrong with either file. They are simply built from different ingredients, and knowing what those ingredients are is the quickest way to decide whether a file can be made smaller — and what it will cost you.

What is inside a PDF

A PDF is a collection of numbered objects. Some are tiny: the page tree that lists the pages in order, a dictionary that says how big each page is, the document title. Others carry the real weight:

• Content streams — the drawing instructions for each page: “move here, set this font, show these characters, draw this line.” Text and vector graphics are stored this way, and they compress extremely well.
• Images — photos, scanned pages, logos and screenshots, stored as pixel data with their own compression (usually JPEG for photos, lossless Flate for flat graphics).
• Fonts — so that the document looks the same on a computer that does not have the font installed. A full font file can be several hundred kilobytes; a subset containing only the characters actually used is often a few kilobytes.
• Extras — attachments, embedded colour profiles, form fields, comments, thumbnails, and metadata such as the XMP block.

In almost every large PDF, one of these categories dominates. Usually it is images.

The usual suspects

Scans at high resolution

A scanner set to 300 dots per inch in colour produces an image of about 2,480 × 3,508 pixels for an A4 page — roughly 8.7 million pixels, or around 26 MB before compression. JPEG brings that down a great deal, but a multi-page colour scan still easily reaches several megabytes per page. Scanning text documents in colour when greyscale would do, or at 600 DPI when 200 would be readable, multiplies the size for no visible gain on screen.

Photos pasted at full size

When a 12-megapixel phone photo is inserted into a slide or document and shrunk to the size of a postage stamp, many programs keep the full-resolution original inside the file. The PDF exported from it then carries every one of those pixels, even though the reader only ever sees a small picture.

Fonts embedded in full

Some export settings embed entire font files rather than subsets. With several typefaces, especially ones with large character sets such as Chinese, Japanese or Korean fonts, this can add megabytes.

Incremental saves

PDF allows a file to be updated by appending changes to the end rather than rewriting it. Each time some editors save, the old version of an edited object stays in the file and a new one is added after it. After many rounds of edits and annotations, a file can carry a lot of dead weight that no reader ever displays.

Two kinds of shrinking

Every compression tool, whatever it is called, does some mix of two things.

Lossless restructuring rewrites the file without changing anything you can see. It drops objects that are no longer referenced, writes a fresh cross-reference table instead of a chain of updates, and can pack many small objects into compressed object streams (a feature added in PDF 1.5). On a file that has been edited many times, this can help a lot. On a file that came straight out of a modern exporter, it often saves only a few percent — the file was already tidy. The Compress PDF tool's lossless mode does this.

Lossy reduction throws information away. The most effective form is downsampling images: storing fewer pixels. Because pixel count grows with the square of resolution, halving the DPI leaves a quarter of the pixels. Going from 300 DPI to 150 DPI keeps 25% of the image data, and the result usually still prints acceptably. The other lever is JPEG quality: lower quality settings discard fine detail and introduce the blocky “artefacts” you see around sharp edges.

A full desktop PDF editor can downsample each image while leaving text and vector graphics untouched. PDFNova's strong mode takes a simpler, blunter approach that works entirely in the browser: it redraws each whole page as one JPEG picture at the resolution and quality you pick. For scanned documents — which are already pictures of pages — this is close to what a dedicated tool would do. For documents with real text, it has serious costs, described next.

What you give up

Turning a text page into an image means the text stops being text. You can no longer select it, search it, copy it, or have it read aloud by a screen reader. Links stop working. Form fields become flat pictures of form fields. Small print may blur. And for a text-heavy document, the image version is frequently bigger than the original, because a paragraph of text described as characters is far more compact than the same paragraph as pixels.

That is why it is worth looking at a size comparison before keeping any compressed file. If the output is bigger, or only a little smaller, keep the original.

A sensible order of attack

1. Fix it at the source if you can. Re-export from the original program with a “smallest file” or “web” preset, which usually downsamples images and subsets fonts properly. Re-scan text documents in greyscale at 150–200 DPI.
2. Try a lossless re-save. It is risk-free. If the file has been edited many times, it may be enough.
3. Remove what you do not need. Deleting unneeded pages, or splitting a large file into chapters, reduces size without any quality loss in the pages you keep.
4. Use image-based compression only for image-based documents. Scans, photo collections and slide decks made of screenshots are good candidates. Choose around 100–120 DPI for on-screen reading and 150 DPI if it will be printed.
5. Always keep the original. Lossy compression cannot be undone.

A note on photo-based PDFs

If you are building a PDF from phone photos, size is decided before the PDF exists. A modern phone photo is often 3–5 MB, and an image-to-PDF converter that keeps images unchanged will produce a PDF roughly equal to the sum of the photos. Cropping to the page and lowering the camera resolution, or running the finished file through strong compression, both help. The guide on making a clean PDF from phone photos covers this in detail.

Summary

Large PDFs are nearly always large because of images, occasionally because of fonts or accumulated edits. Lossless restructuring is safe but modest. Real reductions come from storing fewer pixels, which is ideal for scans and harmful for text documents. Compare sizes, keep the original, and pick the method that matches what the file actually contains.