Compress PDF
Reduce PDF file size while maintaining quality.
Drop the PDF you want to compress
or click to browse
Choose File
Compressing...
Reduce PDF file size while maintaining quality.
or click to browse
Your file is never uploaded. Compression runs on your own machine, which is why a confidential contract or a bank statement is safe to put through it.
The six levels are not a vague quality slider. Each one is a specific instruction about what to touch and how hard.
Re-saves the document without altering a single image: object streams are optimised and metadata is stripped. Typically 5–20% smaller. Choose it when the file must be bit-for-bit faithful, such as an archival scan or a document heading for print.
Only genuinely oversized photos get resized, and they are re-encoded at high quality. The saving is modest and the result is hard to distinguish from the original. Good for portfolios and anything with photography in it.
A real reduction with quality that still holds up on a large screen. A sensible pick for client-facing documents.
The default, and the one to reach for when you have no particular reason to choose otherwise. 1600px is comfortably more than a page needs for screen reading or office printing, and JPEG 70 is where most people stop being able to tell. This is the level that gets email attachments under the limit.
The smallest general-purpose output. Photographs soften visibly if you look closely; text and line art do not, because neither is touched. Use it for upload portals with hard size caps.
Built for pages that are photographs of paper. It separates the sharp text layer from the paper background and compresses each in the way that suits it — the MRC technique behind those suspiciously small 100KB scans. Reductions of around 10× are normal, and text comes out sharper than a plain JPEG at the same size, not softer. Pages that already carry a real text layer or a genuine photograph are detected and left alone, so it is safe to run over a mixed document.
Almost all the weight in a typical PDF is in its images. A 12MB report is rarely 12MB of writing — it is a handful of screenshots pasted in at full camera resolution, or thirty pages photographed at 12 megapixels each. So that is where the work happens: every embedded image is examined, oversized ones are scaled down to the cap for your chosen level, and each is re-encoded at the matching quality before being spliced back into the file.
Everything else is deliberately left untouched. Text is never rasterised. Fonts stay embedded, characters stay selectable, copy-and-paste keeps working, and search still finds words. Vector art — logos, charts, tables, signatures drawn as paths — stays vector and stays crisp at any zoom. Form fields keep working.
This is worth spelling out because the crude approach to PDF compression is to render each page to a JPEG and rebuild the document from those pictures. It reports impressive percentages, destroys the text layer, makes the document unsearchable, and on text-heavy files often produces a larger result than it started with. That approach is not used here.
Sometimes a file barely moves, and there is always a specific reason.
It was already optimised. Exports from InDesign, LaTeX, or a well-configured print driver often arrive close to optimal. There is genuinely little left to remove, and the tool tells you so rather than pretending otherwise.
It is all text and vectors. A 40-page contract with no images has nothing image-shaped to shrink. Lossless mode will strip metadata and tidy the object streams for a few percent, which is the honest ceiling on that kind of file.
It is a scan, and you used a general level. This is the common one. Scans respond enormously to Scanned Document mode and only a little to the general levels. If a scanned file came back roughly the same size, run it again on that setting.
The bulk is not images at all. Embedded fonts, attachments, or an unusually large form definition can dominate a file. Edit Metadata shows you what a document is carrying, and Flatten PDF bakes down form fields and annotations when they are no longer needed.
Most size caps in the wild are familiar numbers: 25MB for Gmail attachments, 10MB or 5MB for government and university upload portals, and 2MB for some job-application forms. A practical route to any of them:
Yes, on every level. Text and fonts are never re-rendered — only embedded images are recompressed. Copy-and-paste, in-document search and screen readers all behave exactly as they did before.
Check what is inside it. A text-only document has no images to shrink, so Lossless mode's few percent is the real ceiling. A scan should be run through Scanned Document mode, which typically achieves around 10× where the general levels manage very little. A file exported by professional layout software may simply already be optimised — the result message will say when nothing recompressible was found.
Scanned Document, without hesitation. It handles photographed and scanned pages by treating text and paper background separately, so the writing comes out sharp rather than smeared, and the file typically lands around a tenth of its original size. Any pages that already have a text layer are detected and passed through untouched.
Lossless mode is, by definition — no image data changes. The other levels re-encode photographs, and that step cannot be undone, which is why the original file on your disk is never modified. Keep it until you have looked at the output.
It depends entirely on what is inside. Image-heavy documents on Balanced commonly land at 30–60% of their original size. Scans on Scanned Document mode routinely reach a tenth. Text-only documents move by a few percent, because there is nothing else to remove. The result line reports the actual figure and how many images were recompressed.
One file per run keeps the memory footprint small enough that large documents work on a phone. For a batch, run them one after another — each takes seconds — or Merge PDF them into a single document first and compress that in one pass, which is often what you wanted anyway.
Rewriting a document invalidates any cryptographic signature it carries, which is true of every PDF editor — so compress first, then sign with Sign PDF. For a password-protected file, remove the password with Protect PDF, compress, then apply it again.
No. The PDF is decoded, its images recompressed, and the document rewritten entirely inside this browser tab. Nothing is transmitted, nothing is stored, and closing the tab disposes of everything. Disconnect from the network before you start if you want to confirm it — the tool still works.
Privacy: Every byte of your PDF stays in your browser. Compression runs locally with PDF.js and pdf-lib. No uploads, no copies.