What Compress PDF decides for each image

When an image is resized and when it isn't

An image in a PDF is stored once and can be drawn in several places. A company logo in the header of a 40-page report is often one image drawn 40 times. The tool measures every place an image is drawn and works from the largest one, so the logo is sized for its biggest appearance, not its smallest.

From that size it works out the image's current resolution and compares it with your setting. An image is only scaled down when it's more than 15% sharper than the target. At the default 144 ppi, that means anything up to about 165 ppi keeps its pixel size and is just re-saved at the chosen quality.

Two more rules apply:

• Images under 64 by 64 pixels, such as icons, are left alone.
• A JPEG is replaced when the new copy is at least 3% smaller. An image stored without loss (the kind used for screenshots and diagrams) is turned into a JPEG only if that saves at least 20%.

Grayscale here only affects images

"Grayscale" under Image conversion works on the images the tool re-saves. Coloured text, coloured table borders and vector charts stay in colour. So do the images it skips for the reasons above, and image types it never touches, such as CMYK photos from print-ready files.

If you need the whole document in shades of grey, for a print shop or a submission rule, use the Grayscale tool instead, then compress the result.

What the clean-up step throws away

These parts are always removed, whatever settings you choose:

1. Page thumbnails, small previews some programs store for each page.
2. Private application data. Adobe Illustrator, for example, can keep a full copy of its own editable artwork inside a PDF. Removing it often saves a lot, but Illustrator can then only open the file as a plain PDF, so keep the original if you still edit it there.
3. Objects nothing refers to any more, such as leftovers from earlier edits.

Data stored without any compression is packed, where that helps, with the standard lossless method used inside PDFs, and the internal structure is saved in a more compact form. None of this changes how a page looks.

"Remove document metadata" deletes both places where a PDF keeps its properties: the basic information list (title, author, creating program) and the XML metadata block that newer software reads.

Font optimization and when to skip it

"Optimize" only touches one kind of font: fully embedded TrueType fonts of the type office suites usually write. It removes the shapes of characters the document never uses and stores identical font files once. Fonts used by fillable form fields are left alone, and the text on the page stays exactly as it was.

The catch is editing later. If someone opens the compressed file in an editor and types new words in that same font, characters that were removed can come out blank or in a fallback font. If it's a draft that will be edited again, leave Fonts on "Leave unchanged".

To compress the same way every time, for example grayscale then compress, set it up once in Workflows.