Compress a PDF without losing quality

A PDF is a container. Text, fonts, and images sit inside it, sometimes with extra copies of objects an export never cleaned up. Shrinking the container is not one dial from pretty to ugly. Some passes remove waste and leave the page looking the same. Some passes throw information away. Lossless PDF compression, in the strict sense, is only the first kind: the bytes get smaller and the page can still be rebuilt as the same text and the same image samples.

People rarely mean that strict sense. They mean the file should still be readable, the text should still select, and the photographs should not turn to mud. That is a size you choose, plus a refusal to rasterize until you say so. pdfly is built as that second thing. It will not claim a percentage off every file, because a text export and a camera roll do not shrink by the same amount.

If the PDF is large because of images, fonts, or a scan, the cause is on Why is my PDF so large. This page is about which passes touch the file, and which ones wait.

A file already under the target is returned as it is. Nothing is recompressed for sport. If the file is over the target, gentler passes run and stop at the first size that fits. They clean structure, subset fonts so unused glyphs can leave, recompress images that were stored wastefully, and refry the PDF. Text stays selectable through those passes. You do not pick them one by one. The size you set is what stops them.

Rasterizing is last, and only if you allow it. It turns pages into images. That can make a stubborn file smaller, and it removes the text layer. Words on a rasterized page cannot be selected or searched. A signature drawn as vectors becomes part of the picture. Leave the option off when a clerk, a reader, or a future you has to copy a line out of the PDF.

Acrobat's Reduce File Size command is a preset, not a target and not a promise about text. It can recompress images harder than you wanted, and it still may miss the number you needed. Use the size on this page, and leave rasterizing off, when the requirement is a smaller file that still behaves like a document.

The tool opens at 200 KB. 500 KB, 1 MB, and 2 MB sit beside it. Custom takes a size in KB or MB inside the published range. A looser target is the gentlest request. The passes stop as soon as the file fits, so a 2 MB target does not keep grinding a file that already fits at 2 MB. A tight target asks more. 200 KB on a photo-heavy PDF may need image recompression. It may not need rasterizing. You see the result, and you can run the original again if you want a different size.

The file you download is at or under the size you chose. If that size cannot be reached with the passes you allowed, pdfly tells you the smallest size it got to and does not give you a file that is still over the target. That report is the quality boundary for this file. Turning rasterizing on moves the boundary. It does not invent detail the scan never had.

You can upload a PDF up to 100 MB. More than 500 pages is rejected, and so is a password-protected PDF or a file pdfly cannot read. The target has to sit between 8 KB and 100 MB. The upload and the file pdfly makes are deleted within 1 hour. Download the result before you close the page. Keep the original if you might want another target.

A text export, a contract with a few logos, or a slide deck that embedded a whole font for a handful of words often shrinks without looking different. Subsetting the font and dropping unused objects is invisible on the page. Those files are the ones lossless PDF compression actually describes.

Photographs are already lossy in almost every camera PDF. Recompressing them is a second loss, aimed at the target. A mild target may be hard to see. A harsh target will look soft. That is the trade the size is making. Rasterizing on top of that is a third step, and it is optional.

A scan is images of pages. There may be no text layer to protect, unless OCR added one. Shrinking the images is what changes the size. If an OCR layer is present and you need it, leave rasterizing off so that layer is not thrown away with the first attempt. If the scan was never text, rasterizing and image recompression are close cousins. The selectable-text question is the one that decides.

1. Drop the PDF in the tool above.
2. Keep 200 KB, or choose a looser chip when the file only has to clear a larger line.
3. Leave rasterizing off. That is the whole of the quality choice.
4. Press Compress PDF. If the target cannot be reached, read the floor before you allow rasterizing.

A second run should start from the original, not from a rasterized download. The rasterized file has already discarded the text. The original still has it.