Where the size of a PDF comes from
A page of plain text takes only a few kilobytes. What makes a PDF big is almost always the pictures inside it: scanned pages (each one a large photograph of paper), photos placed in documents, and slides or reports that carry pictures far larger than they are shown. Fonts add a little, and some programs save page content with no compression at all.
That is why one PDF of a hundred pages of text is smaller than another of three scanned pages. It also tells you what a compressor can do: it can shrink the pictures and pack the rest more tightly, but it cannot make text smaller than text is.
What this tool does to the pictures
- It looks at how big each picture is shown. The tool reads the pages (and the forms and annotations on them) to learn the size every picture is drawn at, then counts the pixels per inch (dpi) it has there. A 3,000 pixel logo drawn 80 points wide has 2,700 dpi, far more than any screen or printer can show, so it is scaled down. A scan at 200 dpi becomes 150 dpi at Balanced. A picture whose use cannot be measured, for example one used in a pattern or a stamp, keeps all its pixels and is only recompressed, and the result says so.
- Photographs become JPEG at the lowest quality that still looks right. Each picture, grey ones too, is compared with the original, using the same measure of similarity as the Image Compressor, and the lowest quality that keeps its score is used.
- Screenshots, charts and logos get a palette. A picture with few colours is stored as an indexed picture with at most 256 colours, the way PNG does, which is far smaller and keeps edges sharp. If a JPEG of the same similarity is clearly smaller, that is used instead.
- Colour profiles, layers and tags stay. A picture that was in AdobeRGB or Display P3 is still in that colour space after it is made smaller, so its colours do not shift. A picture on a hidden layer stays hidden, and its link into the document structure is kept.
- Transparency keeps its edges. The soft masks that give pictures see-through edges are recompressed without loss and scaled together with their picture, so shadows and soft edges do not get dark outlines. A picture whose edge is blended with a background colour (a matte) is recompressed but not made smaller.
- Some pictures are left alone. Black and white scans, CMYK pictures and pictures with unusual settings cannot be recompressed safely, and pictures that would not get smaller are kept as they are.
Everything else is repacked: page content that was saved without compression is compressed, content that was compressed lightly is packed again at the strongest setting, thumbnails and private data of the program that made the file are dropped, and the structure is written more compactly. A file that was PDF 2.0 stays 2.0. The result is read back to check it before you get it.
Choosing a level
- High quality keeps pictures at up to 220 dpi and a high JPEG quality. Use it for documents that will be printed or zoomed in on.
- Balanced keeps up to 150 dpi, which is sharp on screens and fine for most printing of text documents with pictures. It is the default and the right choice for email and uploads.
- Smallest keeps up to 110 dpi and a lower quality. Pictures are still clear on a screen but look soft when zoomed in or printed.
- Stay under a file size starts at your chosen level. If the file is still too big, it looks for the highest picture resolution at which the file fits, down to 56 dpi, and lands just under your limit. In our test, a 4.8 MB PDF of six full page photos reached 983 KB for a 1 MB limit (pictures at about 127 dpi), 460 KB for 500 KB (about 101 dpi) and 191 KB for 200 KB (about 69 dpi). A limit counts a kilobyte as 1,000 bytes, and the result says whether it was met.
Removing the title, author and other metadata
A PDF records who made it and with what in several places: the document information (title, author, subject, keywords, program, dates), a second copy of it as XMP, the metadata of single pages and pictures, and sometimes old copies that earlier saves left in the file. Switch on Remove title, author and other metadata and the tool deletes all of these instead of hiding them, then opens the new file to check that none is left, and tells you what it found.
Names written in the pages themselves, in comments, in form fields or inside the definition of an XFA form are not metadata and stay. The option also works when compressing would save almost nothing: you then get the rewritten file with the metadata gone, and a note if it is a little bigger.
What compression cannot do
- Make a text PDF much smaller. If the pictures are already small, there is little left to remove.
- Open protected PDFs. A PDF that needs a password cannot be read, so it cannot be compressed. Remove the password in the program that made it first.
- Keep a digital signature valid. Any change breaks the signature. The compressed copy is a new file, and the tool tells you if the original was signed or had rights that let Adobe Reader save a filled form.
- Promise that PDF/A still holds. A file that says it follows the PDF/A archive standard gets a note, because recompressing pictures can break the claim. Check the new file with a PDF/A validator if you need it.
- Turn pages into pictures. Some tools do that for the smallest possible size, but the text then stops being text. This tool never does.