A PDF exported from a design tool can weigh 25 MB for one page. The weight is rarely text. It sits in images: 2048 x 2048 pixel textures drawn three inches wide, each with a soft mask for its transparent edges. To compress a PDF like that, this tool measures how large every image appears on the page and cuts its pixels to a target of pixels per inch (ppi).
Compress PDF images by the size they are drawn at
The tool reads the page content, follows the transformation of each image (also inside nested form objects) and takes the largest size at which the image is drawn. A 2048 x 2048 image drawn 228 x 228 points wide, that is 3.17 inches, needs 475 x 475 pixels for 150 ppi. That is 5.4 percent of the original pixels. The reduction is an area average, so it adds no JPEG artifacts.
An image that appears on two pages keeps the resolution of the larger use. An image whose size cannot be measured is limited by its long edge instead: 1800 px in Balanced.
Lossless first, JPEG only when it wins clearly
Many exported images use the lossless Flate format with 8 bits in RGB, gray or an ICC-based RGB profile. After the resolution step the tool stores such an image lossless again and also tries a JPEG version. The JPEG replaces the lossless version only when it is smaller than 60 percent of it, so a photo and a flat logo in one file can end up in different formats. For an ICC-based image the colour profile entry stays as it was. A gray image saved as JPEG becomes RGB.
Soft masks, the transparency of an image, are reduced to the same new pixel size. They stay lossless gray, because a mask must not be a JPEG.
Steps from the drop field to the download
- Drop one or more PDFs on the field or press Enter on it to open the file picker. Each file gets a row with a short analysis: how many images it holds, how many the tool can reduce and how many are drawn at more than the target ppi. An encrypted file asks for its password in its row. Use Add more files to add PDFs later.
- Pick a Level under Settings. It applies to all files. Each level shows its ppi and JPEG quality, and the analysis of every file updates.
- Tick Remove metadata if title, author and dates should not travel with the file.
- Press Compress PDF. The tool works on one file after another. A bar shows the whole batch, and the row of the current file counts images and then masks. A file that fails does not stop the others. Cancel stops after the current step and keeps the files that are done.
- Compare the sizes in the result card, then press Download PDF in a row, or Download all (ZIP) when two or more files got smaller.
Light, Balanced, Strong: the numbers behind the levels
Light keeps 220 ppi at JPEG quality 85 and limits images of unknown size to 2400 px. Balanced, the default, keeps 150 ppi at quality 70 and 1800 px. Strong keeps 110 ppi at quality 55 and 1200 px. Under Custom you set 72 to 400 ppi and a quality from 30 to 95. The fallback edge is then twelve times the ppi. Images that are drawn at less than the limit are never enlarged.
What stays as it is, and what cannot shrink
Text and vector shapes are not changed. In tests with a generated PDF, the page count, page sizes, page content streams, fonts, link annotations and form fields were identical after a run. A PDF that is mostly text therefore gains little: only the packed object streams help.
The details of each file count skipped images by reason: indexed, CMYK and other colour spaces, JPEG images with an ICC profile, hard masks and stencil masks, Decode arrays, bit depths other than 8, formats such as JBIG2, JPEG 2000 and CCITT, and images above 100 megapixels. An image that would not get smaller stays as it is. When the whole file is not smaller, the tool says so and offers no download.
Passwords, memory and the network
An encrypted PDF asks for its password, which stays in memory while the file is in the list. The new file is not encrypted and has none of the permission limits of the source. The browser reads the file into memory and all work happens there. Nothing is uploaded. The only request fetches the PDF library script from this site when you open the first file. A PDF can be up to 300 MB, and a batch up to 50 files and 1 GB. The tool reads and processes the files one after another, so only one file is in memory at a time. A lower resolution cannot be undone in the new file, so keep your original, which the tool never changes.