How to Compress a PDF Without Wrecking the Quality

Someone sends you a 60 MB PDF. It is eleven pages. Your mail server rejects anything over 25 MB, the client portal caps at 10 MB, and the deadline was an hour ago. This is one of the most common file problems in office work, and it is almost always caused by the same thing.

Why your PDF is enormous

A page of text in a PDF is tiny. Text is stored as characters plus a reference to a font, and even an embedded font subset is usually well under 100 KB for the whole document. Ten pages of dense writing might total 200 KB. So when a PDF is 60 MB, essentially all of that weight is coming from somewhere else. In descending order of how often it is the culprit:

1. 1Photographs and screenshots pasted at full resolution. A modern phone photo is 4000×3000 pixels. Dropped into a Word document and printed to PDF, it is displayed at maybe 6 inches wide — around 600 pixels at typical screen resolution — but all 12 million original pixels are still stored in the file.
2. 2Scans. A scanner set to 600 DPI in colour produces roughly 100 MB of raw pixels per A4 page before compression. At 150 DPI grayscale, the same page is about 2 MB. The visible difference on screen is close to nothing.
3. 3Fonts embedded in full rather than subset. A complete CJK font can be 15–20 MB on its own. Most generators subset properly; some do not.
4. 4Leftover invisible content. Cropped images keep their cropped-away pixels. Hidden layers, deleted-but-not-purged objects, and old revisions from incremental saves all still occupy bytes.
5. 5Uncompressed object streams. Real, but usually worth only a few percent — this is the thing generic "optimisers" fiddle with when they cannot touch the images.

The two ways to make a PDF smaller

Downsampling: keep the text, shrink the pictures

This is what you want in almost every case. The tool walks through every image inside the document, works out the size it is actually displayed at on the page, and re-encodes it at a resolution appropriate to that size. A 4000-pixel-wide photo shown at 6 inches gets rewritten at around 900 pixels for 150 DPI output. Text stays text. Vector graphics stay vector. Links and bookmarks survive. The document is still selectable, searchable, and copy-pasteable.

That is how the Compress PDF tool here works by default: it decodes each embedded image, scales it to the target DPI, and re-encodes it as a JPEG, keeping the original whenever the recompressed version would not be meaningfully smaller. Typical results on image-heavy documents are 70–90% reduction with no difference you can see at 100% zoom.

Rasterising: flatten every page to a picture

The nuclear option. Each page is rendered to a single image and that image becomes the page. It compresses brutally well and it is sometimes the only thing that works — particularly on scans that were saved as thousands of tiny image fragments, which defeat per-image optimisation.

The cost is permanent and severe: your text becomes pixels. Nobody can search the document, select a sentence, or copy an address out of it. Screen readers cannot read it. Reach for this only when downsampling has failed and the recipient just needs to look at the pages.

Choosing DPI and quality: a cheat sheet

DPI is the resolution images are resampled to, relative to their printed size on the page. JPEG quality is how aggressively the encoder discards detail within that resolution. These are the settings worth knowing:

Two extra levers are worth knowing about. Grayscale discards the colour channels and keeps only luminance; on a scanned black-ink document that is a straightforward 30–50% saving with no meaningful loss, because there was no useful colour information to begin with. And on documents that are already reasonably sized, [web optimisation](/en/web-optimize-pdf/) restructures the file so a browser can display page one before the rest has downloaded — it barely changes the file size but makes a large document feel dramatically faster to open online.

A practical workflow for hitting a size limit

1. 1Start at 150 DPI, quality 0.8

   This is the sweet spot and it clears most limits on its own. Check the result before doing anything more drastic — people routinely over-compress a file that was already small enough.

2. 2Open the output and look at the worst image

   Find the most detailed photo or the smallest text inside a screenshot, and zoom to 100%. If that looks acceptable, the whole document is acceptable. Judging compression from a thumbnail is how you end up sending something unreadable.

3. 3Still too big? Drop to 96 DPI, and turn on grayscale if it is a scan

   This combination will get a scanned document down by an order of magnitude. It is the right setting for a signed form you are emailing back to someone.

4. 4Still too big? Remove what nobody needs

   A 200-page appendix compressed hard is still large. Use Extract PDF pages to send only the relevant section, or Remove PDF pages to drop the boilerplate. Cutting content beats crushing pixels.

5. 5Last resort: rasterise

   Accept the loss of searchable text, or split the document into parts with Split PDF and send it in two messages. Splitting is usually the better answer.

Mistakes that make compression worse

• Compressing twice. JPEG is lossy and generational — re-encoding an already-compressed image adds new artefacts on top of the old ones without saving much. Always compress from the original, not from a previously compressed copy.
• Zipping a PDF. PDFs already use deflate internally, and their bulk is JPEG data, which is incompressible. A ZIP typically saves 2–5% and adds a step for the recipient.
• Blaming the page count. A 400-page text document can be 3 MB; a 2-page brochure can be 80 MB. Length is a poor predictor of size — image content is the predictor.
• Using "Save As" in a word processor and hoping. Most export dialogs have an image-quality setting buried in the options. Finding it before you export saves a compression round trip entirely.
• Assuming the scanner default was sensible. Many scanners ship set to 600 DPI colour. Setting them to 200 DPI grayscale for text documents fixes the problem at the source, permanently.

When you should not compress at all

Some documents are supposed to be big. Archival scans held as a legal record, artwork proofs going to a printer, engineering drawings where a hairline is meaningful, and anything you may need to re-derive a high-resolution version from later should be left alone. Compression is lossy: the discarded pixels are gone, and no tool can invent them back. Keep the original of anything that matters and compress a copy.