Why PDFs Get Big and How to Shrink Them
A PDF that is too big to email is almost never heavy because of its text. A page of text with embedded fonts takes a few tens of kilobytes; a single photo or scanned page can take several megabytes. Knowing where the size comes from tells you what can be shrunk, what cannot, and what you give up in the process. This guide walks through the numbers and the trade-offs.
Where the size comes from
Text is stored as characters plus the font needed to draw them, and fonts are usually embedded only as the subset of letters the document uses. Vector drawings such as charts and diagrams are stored as shapes. Both are small. Images are stored as pixels, and pixels add up quickly: a phone photo placed in a document keeps its full camera resolution, often 12 megapixels or more, even if it is printed the size of a stamp.
Scans are the other heavy case. A scanner set to 300 DPI turns each A4 page into an image of 2480 × 3508 pixels, about 8.7 million pixels, which is 26 MB uncompressed in colour. Compression brings that down to somewhere between a few hundred kilobytes and a few megabytes per page, depending on the content, which is why a 30-page scanned contract can easily pass 20 MB.
Resolution and quality: the two levers
Resolution decides how many pixels an image keeps. Halving it in both directions keeps a quarter of the pixels: the same A4 page at 150 DPI is 1240 × 1754, about 2.2 million pixels. For reading on a screen, 150 DPI is usually plenty, and text in a scan stays legible at that level; 200 to 300 DPI is worth keeping for documents that will be printed or that contain small print.
JPEG quality decides how much detail is discarded within those pixels. Quality around 70 to 80 typically cuts a photo's size by more than half with little visible change, while lower settings start to show blocky artefacts, especially around text and sharp edges. A good compressor lets you combine the two: downsample oversized images and re-encode them, and leave text and vector content untouched so the document still searches and selects normally.
What compression cannot fix
Some PDFs barely shrink. A document that is mostly text and vector graphics has little to remove; one whose images are already modest JPEGs has little left to squeeze; and images stored as transparency masks, CMYK, fax or JPEG 2000 are often best passed through unchanged, because re-encoding them risks changing colours or breaking the page. If a compressor reports that it left images alone, that is usually the reason.
A scanned page with no text layer is also still an image after compression: it cannot be searched or copied. That needs OCR, a separate step that recognises the letters and adds an invisible text layer. Compressing first and running OCR afterwards keeps the file smaller.
Merging, splitting and what gets lost
Merging several PDFs into one usually copies each page's content, fonts and images into a new document. Document-level parts do not come along: bookmarks, form fields, accessibility tags and digital signatures are typically lost, and a signature is invalid by definition once the bytes change. If any of those matter, keep the original files as well.
Splitting or extracting pages works the same way in reverse and does not reduce the size of the pages themselves. Password-protected files have to be unlocked in a PDF reader first; a tool that pretends to process them without decrypting produces a file no reader can open.
Making a small, readable PDF
When creating a PDF from photos, resize them before or during conversion rather than embedding full camera resolution, and choose a fit that does not enlarge small images: a 200-pixel-wide picture stretched across an A4 page prints at under 30 DPI and looks blurry. For scans, 150 to 200 DPI in greyscale is often a third of the size of 300 DPI colour and is perfectly readable.
Many email services cap attachments at around 20 to 25 MB, including the encoding overhead, so a file just under the limit can still bounce. When a document is still too large, splitting it into parts or sharing a link is more reliable than compressing it until the text becomes hard to read.
- Check whether images or scans make up most of the file
- Downsample to about 150 DPI for screen reading, 200 to 300 DPI for print
- Use JPEG quality around 70 to 80 for photos
- Keep originals if bookmarks, forms or signatures matter
- Compress before running OCR on scans
- Split or share a link rather than crushing text into illegibility