How to Reduce PDF File Size Without Turning Pages Into Pictures

Where the bytes actually are

A PDF is a container, not an image. It holds page descriptions written in a small instruction language - text positions, font references, vector path commands - plus whatever image objects the author placed on those pages. The description is compact by design: it says "draw these glyphs from this font at these coordinates", so the file never stores a picture of the text.

Images are the opposite. Scanned pages and photographs are stored as complete rasters, one value per pixel. A colour scan at 300 dots per inch on A4 is roughly 2480 by 3508 pixels, about 8.7 million pixels a page, and a scanner compresses each page separately, so a fifty-page scan starts life with fifty independent images. Even lossy JPEG compression leaves such pages in the hundreds of kilobytes each.

This is why a 60-page contract written in Word can be smaller than a 3-page scanned invoice: the contract is mostly instructions, the invoice is mostly pixels.

Two different kinds of shrinking

Lossless structural optimisation

This rewrites how the file is organised without touching a single pixel or glyph. Objects that appear many times are stored once and referenced; small objects are packed into compressed object streams; unused entries such as leftover metadata or duplicate resources are dropped; uncompressed streams are deflated. qpdf and Ghostscript's structural settings do this, and it is what Save As in an editor usually performs.

It cannot damage the document: text stays text, images stay bit-identical, and printing or archiving the result is safe. The limit is the size of the prize - a well-built PDF was already optimised when it was created.

Lossy image re-encoding

This targets the images. Each embedded image is decoded, scaled down if it is far larger than the page needs, re-compressed at a lower quality, and written back in place of the original; text, vectors, links and form fields are left alone. This is where the large savings live, and it is the only method that can visibly change how a document looks, so it needs care with small print and fine detail.

A practical rule: try structural optimisation first when the PDF is mostly text and you need an untouched, archivable file; re-encode images when the file is large because of photographs or scans.

What no tool can shrink

Some PDFs come back almost unchanged. It is worth recognising them before blaming the tool.

• Text-only PDFs. A letter, a resume or a report exported from a word processor contains no embedded images. The file may be a few dozen kilobytes to begin with, and rasterising pages would only make it bigger while destroying selectable text.
• Black-and-white fax and bilevel scans. These use CCITT Group 4 or JBIG2 coding, a specialist scheme that stores pure black-and-white pixels extremely efficiently. A browser has no encoder for it, and converting such a page to greyscale JPEG is normally a step backwards in both size and sharpness. Rescanning is the practical fix.
• Raw and palette bitmaps. Images stored with no filter, or as an indexed palette, are often already tiny; converting them to JPEG can make them larger and add colour fringing.
• Images with transparency. Flattening an image that carries an alpha channel would put a visible background where transparency used to be.
• Encrypted PDFs. With password or certificate encryption, the contents cannot be rewritten without the password.
• Already optimised files. If the images inside are already modest, a second pass saves very little.

A good tool says which of these happened instead of reporting a success that did not occur. The same applies to images stored with JPXDecode (JPEG 2000), which browsers rarely support for re-encoding. When a file cannot be made smaller, change the source: scan at a sensible resolution, ask for a different export, or split the document.

Digital signatures and other hard limits

This is not a flaw in a particular program. It is what makes a signature meaningful: it proves the document has not changed since it was signed. Compress a signed contract and send it on, and the recipient's verification fails.

• Passwords. An encrypted PDF has to be unlocked before you can compress it.
• Certification signatures and document timestamps. These are signature variants and carry the same constraint.
• Linearisation. A PDF optimised for fast web viewing stops being linearised once rewritten, so it may open slightly slower over a slow connection.

Hitting a real size limit

Most people are not compressing for its own sake. They are up against a specific number a mail server or portal enforces. Treat the targets below as guidance; the exact requirement always comes from the destination.

If a single pass does not reach the limit, work through these options in order:

1. Compress at the smallest setting, then judge readability at 100 per cent zoom rather than by file size alone.
2. Remove what the destination does not need. Blank pages, cover sheets and unrequested annexes are free savings nobody pays for in quality.
3. Split the document. Many portals accept several files, so a 12 MB report sent as three parts passes a 5 MB limit.
4. Recreate the PDF from the source when possible; exporting with a lower image quality setting often beats post-processing a compressed file.
5. Rescan rather than re-compress. A scan at 200 to 300 dots per inch with the scanner's own compression normally beats shrinking a 600 dpi scan afterwards.

One thing not to do: rasterising pages to hit a size target. Tools that convert every page into an image destroy selectable and searchable text and can produce a file larger than the original.

Compressing in the browser, including in bulk

The PDF compression tool on this site runs entirely inside your browser with the libraries the page has already loaded. Your PDFs are read from disk into the tab, processed there, and saved back to disk. Nothing is uploaded and no account is needed, which makes it suited to contracts, payslips and identity documents - the files most often required to be smaller before sending on.

1. Drop one or more PDFs onto the page, or click to choose them. Batch processing is supported.
2. Pick a compression level: the smallest setting re-encodes images at the lowest quality and smallest dimensions, balanced sits in the middle, and the best quality setting keeps more detail. It applies to every file in the batch.
3. Start the compression. Files are processed one after another with per-file progress, so a large batch does not lock up the tab.
4. Review each result: size before, size after, how much was saved, how many images were re-encoded and how many pages the document has.
5. Download individually, or download the whole batch as one ZIP archive, built in the browser as well.

The tool replaces embedded JPEG images with re-encoded ones and leaves the rest of the file alone, so text stays selectable and searchable, links keep working and page dimensions do not change. It also reports the situations it cannot help with:

• There are no embedded images, so the file is already about as small as its content allows.
• The images are stored in a form the browser cannot decode and re-encode, such as raw bitmaps, palette graphics or JPEG 2000.
• The document looks like a black-and-white fax or bilevel scan, which no browser-side tool can re-encode usefully.
• Re-encoding an image would have made it larger, so the original image is kept.
• The file carries a digital signature, which the rewrite would invalidate.
• The PDF is encrypted and needs unlocking first, or exceeds the 150 MB per-file limit.

If the compressed file would not have been smaller, the unchanged original is what you get.

Frequently asked questions

Tools used in this guide

More guides