Tested guides · Original examples
PDF larger after compression: reproduce the result with one scan
Quality is not a compression ratio. Re-encode an already compressed scan to see why size increases and when to keep the original.
Maintained by:Finl Tools · Updated: · Example dates and methods appear below
Open PDF compressionOne input, two independent runs
Input: 256435 bytes. The outputs below are measured results, not estimates.
| Quality setting | Output bytes | Change from input |
|---|---|---|
| 70 | 330527 B | Increase of 28.9% |
| 95 | 587963 B | Increase of 129.3% |
Conclusion: neither run reduced the file size. Keep the input rather than replacing it with a larger output.
Understand the limits of this test
The input is our synthetic scan-q40.pdf: one A4 page already JPEG-reencoded once at quality 40. It is not an original scan master and contains no real business data. This test specifically examines raising quality on an already compressed file.
Use the same input for both runs: quality 70 first, then quality 95, both with PDF output. Do not use the quality 70 output as the second input, which would change the test conditions.
Reproduce the results and check the page count
The input is 256435 bytes. Quality 70 produces 330527 bytes, an increase of 28.9%; quality 95 produces 587963 bytes, an increase of 129.3%. Both remain one page with unchanged A4 dimensions.
Download all three files, check exact byte counts in file properties and compare the same small text and fine lines in a reader. Reader size displays may be rounded; the table below uses actual bytes.
Results depend on the source, encoder and versions. The table is not a rule for all PDFs or a compression-ratio promise. A larger file does not itself prove improved image quality.
Why higher quality cannot recover lost detail
The current tool rewrites PDF content streams and tries to re-encode supported embedded images at the selected quality. Saving an already low-quality JPEG at higher quality can use more bytes to represent existing pixels.
Detail lost during the first lossy encoding does not automatically return when a later quality setting is raised. To improve clarity, return to the original scan rather than repeatedly transcoding the current output.
PDFs can also contain text, vector graphics, fonts and other objects. Not every size change comes from image quality; scan settings are not necessarily suitable for documents with few images.
Choose the next step based on your goal
If the original is already below the receiving limit and clear, use it directly. Keep the original when the output grows; there is no need to replace it just because processing succeeded.
If still over the limit, inspect the PDF structure, then try a lower quality from the original and check small text each time. If size and clarity cannot both be met, remove irrelevant pages or split by chapter if the recipient permits it. Not all upload platforms accept split files.
For documents with signatures, forms, bookmarks or annotations, size and visual comparison do not validate every feature. Keep the original before processing and check these features before delivery.
Download the input and actual outputs
All files are synthetic examples created by this site without real business information. Download them to repeat the operations.
2026-10-08: called this site’s /api/pdf/compress endpoint twice with the same input and PDF output. Environment: pypdf 6.19.0 / PyMuPDF 1.28.2 / Pillow 12.3.0. Reproduction scripts and byte records are saved with the website project.
| File | Actual bytes | Download |
|---|---|---|
| Input: an already compressed scanned page | 256435 B | Download file |
| Actual output at quality 70 | 330527 B | Download file |
| Actual output at quality 95 | 587963 B | Download file |