Three synthetic files, the live compression tool, and the Balanced setting. Two files barely shrank. The text document became 24.51% smaller. Here are the actual sizes, quality checks and files to inspect for yourself.
Commissioned by PlumeDocs, the maker, and prepared and checked with AI assistance. This is a small sample test, not an independent benchmark.
01 / Original samples
One page per file · Balanced
SAMPLE 01
1 page0.059%
smaller · 556 bytes saved
A scanned page with a JPEG image. The image stream stayed unchanged; the file saved 556 bytes.
No extractable text before or after. This was not an OCR test.
SAMPLE 02
1 page0.008%
smaller · 719 bytes saved
A poster with two large raster images. Both image streams stayed unchanged; the file saved 719 bytes.
1,246 extracted text characters, identical before and after.
SAMPLE 03
1 page24.51%
smaller · 649 bytes saved
A text-only page with no embedded images. The file saved 649 bytes while its extracted text stayed identical.
1,063 extracted text characters, identical before and after.
What stayed the same
All three original outputs kept their page count, page geometry and extracted text. Their before-and-after RGB renders were pixel-identical at 150 dpi in Poppler. That is a result for these files and this renderer, not a guarantee for every file or PDF viewer.
Downloads start only when selected. File-size labels use decimal units: 1 kB = 1,000 bytes; 1 MB = 1,000,000 bytes.
02 / A separate experiment
We prepared a second version of the poster by removing a reversible ASCII85 wrapper from its image streams. Its decoded image content, text and page geometry matched the original; its 150 dpi render was also pixel-identical before compression.
That preparation alone made the input 19.99% smaller. We do not attribute that saving to PlumeDocs. The run below begins with the already-prepared file and does not replace the original poster result above.
Prepared poster → Balanced output
7.30 MB → 53.59 kB
7,304,086 bytes → 53,592 bytes
99.27% smaller in this separate case
The two images became JPEGs and went from 1,800 × 1,800 to 1,600 × 1,600 pixels. The output shows subtle smoothing and color changes. It is not pixel-identical or lossless.
The page geometry and all 1,246 extracted text characters stayed the same. At 150 dpi, 26.91% of rendered pixels changed by at least one RGB channel value. This percentage is not a measure of perceived quality loss.
Matching crops from the actual 150 dpi renders, shown scaled to fit your screen. Open each image to inspect the source pixels. The PDF report includes full-page views and additional detail crops.
03 / Method & limits
The inputs are synthetic teaching-style documents, made for this test. They contain no real student records or confidential research. The outputs are the files downloaded from the live PlumeDocs compression tool on 8 September 2026 using Balanced.
Poppler reported a nonfatal Fontconfig configuration warning for text-bearing files. All renders completed and were visually reviewed using the same settings before and after.
In the separate compatibility case, the mean absolute RGB difference was 0.325 on a 0–255 scale; the maximum channel difference was 9. The different outcome suggests encoding affects the result, but this test alone does not establish the product-code cause.
Try a synthetic sample above, then inspect the downloaded output at the size and quality you need.
Open Compress PDF
