Find the heavy pages. Shrink the PDF.
Most of a PDF's size hides in a few pages: usually scans and photos. The map shows where the weight is; pick a level and watch it drop.
A PDF exported from Word stores words as text, a page is a few kilobytes. A scanned page is a photo of paper: millions of pixels, even when it's mostly white. That's why a 10-page scan can outweigh a 300-page report.
Tip: if you still have the original document, export a fresh PDF instead of scanning a printout, it'll be up to 100× smaller and searchable.
Your browser just drew a scanned US Letter page at 300 dpi, paper grain included. It then resampled the page to each resolution and saved it as a JPEG. The sizes below are real bytes from your browser's encoder, not estimates. Click a row to see one inch of the page at that resolution.
Raw RGB is width × height × 3 bytes: what the page costs before any compression. Halving the dpi quarters the pixel count, which is why resampling saves far more than lowering JPEG quality.
At 100% zoom most screens show about 100-150 pixels per inch of page. Office printers resolve around 300. Below 150 dpi, 11-point text starts to lose its serifs, and below 96 dpi letters like e and c begin to merge.
A page's text and vector drawing live in a content stream: a small program of operators such as Tj (show text), l (line to) and re f (fill a rectangle). PDF compresses it with the FlateDecode filter, which is plain zlib. Edit the stream below and your browser compresses it for real.
5 0 obj<< /Length 392 /Filter /FlateDecode >>stream… 392 binary bytes …endstreamendobj
Each stream names its filters in /Filter, and a conforming viewer has to decode all of them. A good compressor picks one per stream: Flate for text and vectors, DCT or JPX for photos, and CCITT or JBIG2 for black-and-white scans.
Most PDF compressors, including Ghostscript's -dPDFSETTINGS, describe their levels by purpose. Underneath, each name is mostly an image resolution target.
Compression is usually one step in getting a document out the door. These handle the rest.
It depends on the content. PDFs with many high-resolution images can often be compressed by 70 to 90%. Text-only PDFs may only reduce by 10 to 30%.
There may be a slight reduction in image quality, but text and document structure are preserved. For most business documents, compressed PDFs look identical to the originals.
Browser-side mode processes supported files locally in your browser. Confirm local mode before using sensitive files.
Password-protected PDFs need to be unlocked before compression. Once the password is removed, the PDF can be compressed normally.
No - the text, layout, fonts, and document structure are preserved. Only the file size changes, primarily through image optimization.