How to Compress a PDF Without Losing Quality (Lossless vs Lossy)

A PDF is a package of objects: page descriptions, fonts, images, annotations and a cross-reference table that tells a reader where each object starts. Text costs almost nothing, because a page of prose is a few kilobytes of character codes plus drawing instructions. Images are different. A single photo placed in a report keeps whatever resolution it had when it was inserted, so a 12-megapixel phone picture shrunk to a thumbnail on the page can still weigh 3–5 MB inside the file.

Scans are the extreme case. Each scanned page is one full-page image, and a color scan at 300 DPI of an A4 sheet is about 2480 × 3508 px. A scanner app that saves those pages with light JPEG compression produces 1–3 MB per page, so a ten-page lease turns into a 20 MB attachment even though it contains maybe 30 KB of actual words.

The remaining weight hides in places you never see. Full embedded fonts can add 200–500 KB each for large CJK or display families, the same logo may be stored once per page, and files edited repeatedly in Acrobat carry incremental updates, where every save appends changes instead of rewriting the file. Attachments, form data and page thumbnails add more.

Lossless steps rewrite the file without changing a single pixel or glyph. Object streams, added in PDF 1.5 in 2003, pack hundreds of small objects into one compressed block instead of storing them as plain text. Deduplication finds identical images and fonts and keeps one copy. Font subsetting keeps only the characters the document uses, and cleanup drops XMP metadata, page thumbnails, orphaned objects and the old revisions left by incremental saves. Together these typically shave 5–30% off a text-heavy file.

Lossy steps change image data, and that is where the large savings live. Downsampling reduces an image to a target resolution, for example from 300 DPI to 150 DPI, which removes three quarters of its pixels. Re-encoding saves it as JPEG at a chosen quality, and converting color scans to grayscale removes two of three color channels. Done with sensible numbers, none of this is visible at normal reading size, which is what people mean by compressing without losing quality.

How much a PDF shrinks depends far more on what it contains than on which tool you use. A report exported from Word or Google Docs is already mostly text with subset fonts, so there is little left to remove. A deck exported from PowerPoint with full-bleed photos, or a stack of phone scans, can lose almost all of its weight because the images were never sized for the page they sit on.

The figures below are typical ranges, not measurements of any one file. Use them to decide whether compression is worth trying at all, and which setting to start with when a document has to fit a specific limit.

Upload portals state a ceiling and reject anything a byte over it. At 200 KB you have room for one to three scanned pages in grayscale at around 150 DPI, which keeps 10 pt print readable; compress a PDF to 200 KB starts at exactly that limit. At 500 KB a color ID scan plus a few pages of grayscale text fits comfortably. At 1 MB and 2 MB the question is usually page count rather than quality, since a ten-page color scan fits at 150 DPI.

Work down in a fixed order so quality goes last. Remove pages the form does not ask for, crop the empty paper around cards and receipts, switch black text to grayscale, then let resolution fall before JPEG quality. Quality below about 40 creates blocky halos around letters, while 120 DPI in grayscale still reads cleanly. A target-size mode does this search for you; compress a PDF to 1 MB handles the most common job and application limit.

Leave a small margin. Some portals count a kilobyte as 1,000 bytes and others as 1,024, so a file reported as 200 KB on your computer can be 204,800 bytes and still be rejected by a form that means 200,000. Aiming for 190 KB, 480 KB or 0.95 MB avoids the argument.

Desktop apps can all make a PDF smaller, but they differ in how much control they give you. Acrobat Pro offers both a one-click compressor and a full optimizer: its PDF Optimizer includes Audit space usage, which lists exactly how many bytes go to images, fonts, content streams and bookmarks, so you can see whether compression will help before you try. Word cannot compress an existing PDF, but it decides how large the PDF will be when you export it.

Preview on a Mac is the tool most people already have, and the one most likely to disappoint. Its built-in Reduce File Size Quartz filter applies one fixed, aggressive setting to every image, with no slider for resolution or quality. Scans come out soft and blotchy, and a text-only PDF can even grow, because the filter rewrites the file without the object-level cleanup a real optimizer performs.

A browser compressor is the shortest path when you are on a work laptop without Acrobat, on a phone, or helping someone over a call. The online PDF compressor on UltKit takes up to 20 PDFs at once, each up to 95 MB, with nothing to install and no account, and every uploaded and processed file is deleted 30 minutes after processing.

The defaults are set for quality: Recommended downsamples only oversized images and leaves text and vectors as they are, while Subset fonts and Remove metadata are already on. Change things only when the result has to fit a number.

A file size is not proof of quality, so open the compressed PDF before you send it. Zoom to 200% on the smallest text in the document, such as footnotes, table figures or the fine print under a signature, and compare it with the same spot in the original. At 200% any JPEG halos, smeared strokes or stair-stepped edges become obvious, while at fit-to-width they hide.

Then check the things compression should never touch. Text you could select and copy before must still be selectable, because a tool that rasterizes pages into pictures has not compressed the document; it has replaced it. Links, bookmarks and form fields should still work, and the page count should match.

Some files are already as small as they can honestly get. A 300-page text manual with subset fonts, an engineering drawing made of vector paths, or a black-and-white scan stored with JBIG2 has no oversized images to shrink, so every compressor returns roughly the same size. Pushing harder only hurts: forcing a vector drawing through rasterization makes it bigger and blurrier at the same time.

When the limit is per file rather than per submission, split the document instead. Cut a long contract into the signature pages and the annexes, a thesis into chapters, or a bank statement bundle into one month per file, then compress each part only if it still needs it. Split PDF cuts by page range or into fixed-size chunks, and the guide to splitting PDF files covers which split suits which portal.

Splitting also helps with email. Gmail caps attachments at 25 MB in total per message, and many company mail servers set lower limits, so two 15 MB parts often travel where a 30 MB file bounces. For scans you are about to make, the better fix is upstream: scan text at 200–300 DPI in grayscale, as described in scanning documents to PDF, and you will rarely need to compress at all.