How to compress a PDF to 1MB, 500KB or 200KB

There is no formula from a setting to a file size

The same quality setting produces wildly different files depending on the picture. A blank scanned page and a photograph of a crowd, at identical settings and identical dimensions, can differ by a factor of twenty. JPEG spends bytes on detail, and how much detail there is depends entirely on what is in the picture.

So nothing can jump straight to your answer. Any tool that hits a size target does it by trying settings, measuring the real bytes, and trying again. If a tool gives you a number without doing that, it is guessing, and you find out at the upload box.

Resolution, not quality

A PDF never records how large a picture appears on the page. It stores the image, then separately stores a matrix that places it. A scanner embeds 300 DPI into a page nobody will print above 150, so two thirds of those pixels are never seen by anyone.

Lowering quality does not find those pixels. It makes every pixel slightly worse, including the surplus ones. Lowering resolution removes them. On the oversampled page below, the image dropped from 2400 x 3200 to 1240 x 1754, the file went from 231 KB to 53 KB, and the page looks the same at reading size because the extra pixels were never visible to begin with.

That is why our compressor lowers resolution before it lowers quality, and keeps the resolution as high as the target allows rather than dropping both together.

The files

• A scanned document4 pages, photographic · 633,988 bytes

  Each page is a 1700 x 2200 image drawn across 595 points, which is 206 DPI.

• A single oversampled page1 page, one large image · 236,753 bytes

  A 2400 x 3200 image drawn across a full A4 page, which is 290 DPI.

• A text-only document55 pages, no images at all · 713,802 bytes

  Exported to PDF by LibreOffice from a plain text file. Not one picture in it.

What each target actually needed

• Scanned document1 MB619 KB, unchanged

  None applied

  It was already under the target. Compressing it would have cost quality to solve nothing.

• Scanned document200 KB127 KB

  220 DPI, quality 95

  The highest settings on offer already cleared 200 KB, so nothing lower was needed.

• Scanned document100 KB99 KB

  220 DPI, quality 85

  Quality came down one step. The resolution never had to move.

• Oversampled pageDefault settings53 KB from 231 KB, 77% smaller

  150 DPI, quality 60

  The image went from 2400 x 3200 to 1240 x 1754, which is exactly 150 DPI for A4.

• Text-only documentDefault settings257 KB from 714 KB, 64% smaller

  No image work at all

  Nothing was re-encoded. The saving is entirely in how the file was packed.

A file that already fits should be handed back untouched

The scanned document asked to fit 1 MB was already 619 KB. The honest answer is to hand it back untouched, and that is what happens now. It is not what happened when we first built this: the search dutifully found the best settings that cleared 1 MB and returned a 127 KB file, five times smaller than asked for and visibly worse, solving a problem that did not exist.

Worth checking on whatever tool you use. Give it a file that already meets the limit and see whether it hands the file back or compresses it anyway.

Text-only PDFs compress too, and most tools refuse them

A 55-page document with no images in it went from 714 KB to 257 KB, 64% smaller, with nothing re-encoded and not one character altered. The saving is entirely in how the file was packed: a word processor often stores page content with no compression at all, and a document assembled by merging several files carries the same embedded font once per source file.

Our own tool used to refuse this file. It said there were no images, so there was nothing to compress, which was wrong on its own terms and stayed wrong for a fortnight. If a compressor tells you a text PDF cannot be made smaller, it is describing its own implementation rather than your file.

When a target is out of reach

Some files will not get there. A 40-page photographic scan asked to fit 100 KB has to give up either the pages or the legibility, and no setting rescues that. A tool should say so and give you the smallest it managed, so you can split the document or go back and ask for a different limit. Handing back something still over the target leaves the upload box to deliver the news.

Method

• Every file was synthetic, generated for this test. No real document was uploaded anywhere, and none of these carry a name or an account number.
• Each was run through the live tool at filetity.com/compress-pdf in Chrome on 3 September 2026, not through a development build.
• Sizes are the bytes of the produced file. Image dimensions were read out of the output PDF’s own dictionaries afterwards, not from what the tool reported.
• Page counts and page geometry were checked before and after with pdf.js, which is a different library from the one that writes the file, so the two are not agreeing with themselves.
• The compressed pages were rendered and compared against the originals to confirm they were the same picture rather than a blank page. Mean brightness after compression was 129.2 against 129.3 before.