Why your PDF is too big, and what actually controls the size

A PDF of five photographed pages can be 400 KB or 40 MB. The difference is almost never the number of pages, and it is almost never something the PDF itself is doing wrong. It is the pictures inside it, and specifically three properties of those pictures.

Knowing which of the three is your problem saves a lot of wasted effort, because the wrong fix barely moves the number.

First: is it a scanned PDF or a text PDF?

These are two completely different objects that share a file extension, and the advice for one is useless for the other.

A text PDF was exported from a word processor, spreadsheet or design program. Its pages are instructions — draw this glyph here, in this font, at this size. You can select the text with your cursor. These files are naturally small: a forty-page report can easily be under 1 MB, and there is usually nothing to optimise.

A scanned PDF is a stack of photographs in a PDF wrapper. Every page is a picture of a page. You cannot select the text, because there is no text — only pixels arranged to look like text. These are the files that come out at 40 MB.

The test takes two seconds: open it and try to select a line of text with your mouse. If a selection highlight appears over the words, it is a text PDF. If you get a rectangle over the whole page or nothing at all, it is scanned, and the rest of this page is about your file.

The three things that control the size

Ranked by how much each one is usually worth:

1. Pixel dimensions — the biggest lever, and the most ignored

This is the one people skip, and it is almost always where the weight is. A modern phone camera produces images of twelve megapixels or more. A sheet of A4 captured at a comfortably readable quality needs roughly two.

The arithmetic is unforgiving, because file size tracks the pixel count and the pixel count is an area. Halving the width and the height of an image removes three quarters of its pixels. Going from 4000 × 3000 to 2000 × 1500 does not halve the file — it typically cuts it to somewhere near a quarter.

Put the other way round: a 12-megapixel photo of a document holds about six times the detail of what you need, and you are paying for all of it on every page.

2. JPEG quality — the fastest lever

Every scanned page inside a PDF is stored with some compression setting, usually JPEG. The quality value is a number from 1 to 100 that decides how much visual detail is thrown away.

The relationship is not linear, and that is what makes it useful. Going from 95% to 85% removes a lot of bytes and almost no visible quality, because the top of the range is spent encoding detail that nobody can see. Going from 60% to 50% removes far fewer bytes and a great deal of visible quality, because by then you are into detail that matters.

For a photographed or scanned document, 70–80% is usually indistinguishable from the original at reading size. Below about 60% you start seeing the characteristic JPEG damage around text: faint halos and smudging at the edges of letters, worst on small type.

3. Scan resolution (DPI) — set once, hard to undo

If you are scanning rather than photographing, the scanner asks for a resolution in dots per inch. This directly sets the pixel dimensions, so it is really lever 1 under a different name — but it is decided before the file exists, which makes it worth getting right.

• 150 DPI — fine for a document that will be read on screen. An A4 page comes out around 1240 × 1754 pixels.
• 300 DPI — the standard for anything that might be printed or needs small print to stay legible. Around 2480 × 3508 for A4, which is four times the pixels of 150 DPI.
• 600 DPI — archival, or documents with very fine detail. Four times again. Rarely necessary, and it is where enormous files come from.

Many office scanners default to 300 or 600 DPI in colour. For a black-and-white text document being emailed, that is several times more than anyone needs.

Colour mode, the hidden multiplier

A scanner set to colour stores three channels per pixel. Set to greyscale it stores one. For a document printed in black ink on white paper, two of those three channels are carrying no information at all.

Switching a text scan from colour to greyscale often cuts the file substantially on its own and costs nothing, because there was no colour in the original. The exception is a document with a coloured letterhead, a stamp, or a signature in blue ink that someone may need to verify — there, keep the colour.

What to change, in order

1. Resize to about 1600–2000 pixels on the long edge. This is enough to read A4 text comfortably and is frequently the difference between 9 MB and 1 MB.
2. Set JPEG quality to 75–80%. Look at the result at 100% zoom before accepting it.
3. Use greyscale if the document has no meaningful colour.
4. Split the document if it is still too large. Many portals that cap a single file at 2 MB will happily take five of them.

Things that do not help

• Zipping it. A PDF full of JPEGs is already compressed. A ZIP will save you a percent or two, and most upload forms reject archives anyway.
• Renaming or re-saving without changing settings. Nothing about the content changes, so nothing about the size does.
• Removing blank pages. Worth doing for tidiness, but a blank page compresses to almost nothing. It is not your problem.
• Printing to PDF again. This usually re-encodes already-compressed images, which makes the quality worse and the file no smaller — sometimes larger.

Why PNG makes PDFs enormous

If your PDF is built from PNG images, that alone may be the whole story. PNG is lossless: it stores every pixel exactly, with no approximation. For a screenshot of flat colour and crisp text that is efficient and correct. For a photograph it is close to the worst possible choice, because photographic noise has no repeating pattern to compress, and a PNG of a photograph can be five to ten times the size of the same image as a JPEG at a quality nobody could tell apart.

Phone screenshots are PNG. Photos of documents, saved out of some apps, end up as PNG. If a PDF of a few pages is unexpectedly huge, check what went into it. The format comparison goes into when each one is the right answer.

The limit may not mean what you think

If you have got a file to what your computer reports as 2.0 MB and the form still refuses it, the two of you may be counting differently. A megabyte is either 1,000,000 or 1,048,576 bytes depending on the convention, a 4.9% gap at MB scale — enough to fail a file that looks exactly at the limit. See KB vs KiB.

Doing it without uploading the document

Worth a word, because the files involved are rarely trivial. The documents people need to shrink are bank statements, payslips, contracts, medical letters and scans of identity documents — and the usual way to shrink them is to upload them to a free website, which means handing a complete personal dossier to a company you have never heard of.

Celveri’s tools run in your browser with no upload endpoint, so the file never leaves your machine: Image Resizer for step 1, Image Compressor for step 2, and Image to PDF to reassemble the pages with a quality setting you control.