By Baylal, founder of PDFcrest · Published November 3, 2026
The short answer
We created a 10-page report with about 300 words per page, then saved it in different ways. The words never changed, only how the pages were stored.
| How the 10 pages are stored | File size |
|---|---|
| Real text (not scanned) | 0.08 MB |
| Black-and-white scan, 300 DPI | 0.74 MB |
| Color scan, 150 DPI | 5.62 MB |
| Grayscale scan, 300 DPI | 18.82 MB |
| Color scan, 300 DPI | 19.17 MB |
| Color scan, 600 DPI | 67.64 MB |
Three things stand out:
Black-and-white was far smaller, but it turns every pixel pure black or white, so it only suits plain text. It would ruin photos and shading.
We ran a 9.59 MB, 5-page version of the 300 DPI scan through our Compress PDF tool at each setting that keeps the pages' text and images in place.
| Setting | Result | Smaller by |
|---|---|---|
| Low Compression | 1.36 MB | 85.8% |
| Recommended | 0.81 MB | 91.6% |
| Extreme Compression | 0.40 MB | 95.9% |
The text stayed easy to read on screen at Recommended and Extreme, the two results whose first page we compared by eye. We didn't test printing, so check a page before you print or send an important file.
Then we ran the text-only file through the Recommended setting. It went from 0.08 MB to 0.07 MB, just 9.7% smaller, because it had no images to shrink. If your PDF is mostly real text and still large, images aren't your problem, so look at fonts next.
Your results will differ. Our test scan had artificial grain added, and grain costs a lot of space until it is smoothed away. A clean scan, or one that is already compressed, may shrink much less.
A PDF can carry the whole font file, or only the letters it actually uses, which is called subsetting. We made a one-page document in Arial and saved it both ways. With the full font it was 442 KB, and with a subset it was 40 KB, about 11 times smaller.
That difference is small next to a scan, but it adds up on short documents, or when one file carries several full fonts. If you export from Word or a design app, look for a setting that embeds only the characters used.
Keep your original until you have checked the smaller copy. Compression can't be undone on the compressed file.
We generated the test files with PyMuPDF on November 3, 2026. Our "scans" are text pages rendered to images, with an off-white tint, soft shading, and random grain to imitate a scanner, then saved as JPEG at quality 90. The black-and-white version was stored as lossless PNG data. They aren't scans of real paper, so treat the numbers as a comparison and not a prediction for your file.
Sizes are shown the way our tool shows them, where 1 MB is 1,048,576 bytes. Compression ran in Chrome on the live site, and the browser recorded no network requests while it ran, so the file wasn't uploaded. We wrote this guide about our own tools, so please test them on your own files too.
Background on how PDFs store images and fonts: ISO 32000, the PDF specification (PDF Association).