Why Is My PDF So Large? Five Causes, Measured Byte by Byte

The short answer

Usually, pictures. A phone photo goes into a PDF at full size, and a scanned page is one large picture. In our tests, PDFs made from a photo or a scan were at least 99.7 percent image data.

Documents with text showed three more causes: structure written loosely, and fonts and pictures copied many times when separate PDFs are merged. Pictures shrank the most under compression, and loose structure went away with a plain re-save. Copied fonts did not shrink at all.

On October 21, 2026, we ran our test files through Compress PDF and other tools on pulsarkit.com. We used Chrome 155 on a 2023 MacBook, and sizes are as the pages showed them. To see where the bytes sit, we also counted every object in each file with our own script.

A photo goes in at full size, and the page decides the rest

Our photo was a 12-megapixel phone shot of a lake from Wikimedia Commons: 4032 × 3024 pixels, 5.1 MB. JPG to PDF placed the JPG in the document as it was. The PDF came out at 5.1 MB, only 1 KB heavier, in under a second. Extract Images from PDF later handed back the same JPG, byte for byte.

Compress PDF limits how many pixels a picture keeps per inch of the page: 150 on Balanced, 96 on Smallest file. With Fit image, the default in JPG to PDF, the page is as large as the photo: 56 by 42 inches. Spread that wide, the photo has only 72 pixels per inch, already under both limits.

So Compress PDF only re-encoded it and kept all 12 million pixels: Balanced gave 1.6 MB and Smallest file 1.2 MB. On an A4 page, the same photo is about 11 inches wide, at 366 pixels per inch. There, Balanced brought it down to 1654 × 1240 pixels and 341 KB, and Smallest file to 101 KB.

Each compression took under three seconds. If the PDF will be printed or read as a document, choose A4 or Letter in JPG to PDF before you compress.

Scans: resolution costs more than color

A scanned page is a picture of the paper. Our test scan had three pages of text in gray at 200 DPI and weighed 2.3 MB. We made three more versions of the same pages: enlarged to 300 DPI, in color with a cream paper tone and color grain, and both.

At 300 DPI, the file weighed 59 to 62 percent more: 3.6 MB in gray and 3.7 MB in color. Color itself added 2 percent on these plain pages. We did not test pages with color photos.

Balanced resamples finer scans to 150 DPI, and all four versions came out between 509 and 539 KB. Smallest file gave 203 to 212 KB at 96 DPI. How text holds up at each level is in our scanned-book test.

A text PDF: tags, fonts and loose packing

Our text document had 20 pages of headings, paragraphs and tables, printed to PDF by Chrome from a web page: 145 KB. Fonts took 37 KB, a quarter of the file. Chrome embedded only the letters in use, as subsets of Georgia and Georgia Bold. The text itself, as drawing instructions for 20 pages, took 31 KB.

The largest share, 58 KB or 40 percent, was tags: 525 small objects that mark headings, paragraphs and table cells for assistive technology. Another 12 KB was the index of the file's 587 objects. Chrome had written both as plain, uncompressed text.

Compress PDF took the document to 85 KB, 41 percent less, while warning that there was little to compress. Remove PDF Metadata, which leaves pages untouched, gave the same 85 KB in under half a second. Both rewrite the file, and the rewrite packs the tags and the index into compressed blocks.

Every tag survived, and the fonts and page content came through byte for byte. On the 2.3 MB scan, the same re-save took off 1 KB: packing helps a file of many small objects, not a file of pictures.

Copies: one logo, forty fonts

Next, we put a 26 KB PNG logo on all 20 pages with Watermark PDF. It was stored once, as 13 KB shared by every page. Watermark PDF rewrites the file too, so the document came to 104 KB, 19 KB more than the plain re-save.

Then we split it into 20 one-page files with Split PDF and joined them again with Merge PDF. The pages were the same, but the file was 833 KB. Each one-page file had carried its own logo and fonts, and the merged file kept all of them: 20 logos and 40 font files.

Compress PDF found the 20 identical logos and kept one, which brought the file down to 584 KB. The fonts stayed: 40 copies of the same three font files, 499 KB, or 85 percent of what was left. Compress PDF does not merge duplicate fonts. When you can, make one PDF from the original documents instead of merging separate PDFs.

What to try, in order

Start by looking inside. Extract Images from PDF hands back the pictures at the size they were stored. With Skip repeated images on, the default, each comes out once. On our merged file, it returned a single logo; on the photo PDF, the original 5.1 MB JPG.

Then run Compress PDF. Pictures shrink the most, repeated pictures are stored once, and a loosely written file is packed on the way. Copied fonts stay, so for a merged file, go back to the source documents if you can.

If the result is still over a size limit, Split by size cuts it into parts under that limit. Its page warns that fonts and shared images are copied into each part. The parts add up to more than the original. Every tool named here runs in your browser tab. They are among more than 200 free tools on PulsarKit, with no account and no subscription.