What Happens When You Click 'Compress'

When your PDF hits our compression pipeline, we first run a structural audit: parsing the cross-reference table, decoding stream dictionaries, and cataloging every XObject, image, and font resource. Next, we walk each page's content stream and classify content into buckets — text, vector paths, raster images, form fields — so each can be optimized with the right algorithm. Raster images go through our perceptual quality analyzer, which chooses between JPEG, JPEG 2000, and JBIG2 encoding based on color depth, dimensions, and detected content type (photo vs.

screenshot vs. scanned text). Fonts get subset to only the glyphs actually used in the document, often shrinking a 400 KB embedded font to 40 KB.

Finally, we recompress all Flate streams with higher compression ratios than most PDF libraries default to, then rewrite the file with a linearized (web-optimized) structure for faster browser rendering. All of this happens in seconds, entirely within our infrastructure, and your original file is purged from memory the moment the compressed version is delivered.