Make PDF files smaller without uploading them anywhere. Text stays selectable and searchable, images get re-encoded, and everything happens inside your browser.
Drop PDFs here, or click to browse
Up to 30 files · no size limit · you can also paste with Ctrl+V
Images wider or taller than this are scaled down.
Sizes each image to how large it is actually drawn on the page. 0 turns this off.
Higher DPI means sharper pages and a bigger file.
Leave empty to just use the mode settings. Runs several passes to land under the target.
Leave empty to keep every page. Dropping pages is often the fastest win.
Shortcuts: Ctrl+O to open files, Ctrl+Enter to compress.
What it does: reduces the file size of a PDF by re-encoding the images embedded inside it. Text, fonts and vector drawings are left untouched, so the compressed file is still selectable, searchable and readable by screen readers.
Where the work happens: in your browser. The PDF is read from your disk into memory and never sent over the network. There is no upload, no server-side copy and no retention window.
Typical results: scanned or photo-heavy PDFs shrink by 70-95%. Text-only PDFs exported from Word or LaTeX are already efficient and typically shrink by 2-15%.
Cost: free, unlimited, no account, no watermark.
The four modes are not just quality presets - two of them use fundamentally different engines. The distinction that matters most is whether your text survives as text.
| Mode | What it does | Text stays selectable | Typical reduction | Best for |
|---|---|---|---|---|
| Lossless | Rewrites the file structure with compressed object streams and drops metadata. No pixel changes. | Yes | 2-15% | Legal and print files that must not change visually |
| Balanced | Re-encodes embedded images with mozjpeg at 75% quality, capped at 2000px. Text and vectors untouched. | Yes | 40-80% | Almost everything - the sensible default |
| Strong | Same engine, images at 55% quality capped at 1400px. | Yes | 60-92% | Email attachment limits, web upload forms |
| Maximum | Renders every page to a JPEG and rebuilds the PDF from those images. | No - text becomes pixels | 50-95% | Scanned documents, or a hard size limit nothing else meets |
If you are unsure, start with Balanced. If the result is still too big, try Strong before reaching for Maximum, because losing selectable text is a real cost - it breaks copy and paste, search, and screen reader access.
Maximum is not automatically the smallest result. It re-renders every page at a fixed resolution, so on a plain document it can easily produce a larger file than Strong while also destroying the text layer. It earns its place on scans and dense, graphic-heavy layouts. Compare the two before you settle - the tool reports the exact before and after size for each.
Almost all PDF bloat is images. A page of text is a few kilobytes of glyph references; a single phone photo dropped into that page is three megabytes. Scanners are the worst offender, because they save every page as a full-resolution colour image even when the page is black text on white paper.
That is why this tool targets images specifically rather than blindly rasterising the document. It walks the PDF object graph, finds each embedded image, decodes it, resizes and re-encodes it as a JPEG, and writes it back into the same slot in the file. The page content stream, the fonts and the vector graphics are never touched.
Some images are deliberately left alone, because re-encoding them would risk corrupting the page:
A file that is mostly made of these will barely shrink. That is not a failure, it means there was nothing safe left to reclaim.
Your PDF is never uploaded. Compression happens with JavaScript running on your own machine, using the browser's built-in image decoder and canvas encoder. There is no request that carries your file, so there is no server-side copy to leak, subpoena or delete.
That matters for the documents people most often need to shrink: signed contracts, invoices, passports and visa paperwork, medical scans, tax returns and bank statements. With a conventional online compressor, all of those pass through someone else's infrastructure. Here they do not.
The page does load Google Analytics, Microsoft Clarity and Google AdSense, which see page views and interaction events in the normal way. They do not receive your file contents, because your file contents never enter a network request.
The JPEG encoder - mozjpeg compiled to WebAssembly - is served from this site rather than a third party, so using the tool does not announce itself to anyone else. The PDF parser and page renderer are still fetched from a public CDN the first time you visit. All of it is program code travelling towards you, not data travelling away from you, and once loaded it runs entirely on your machine.
No. The compressor runs entirely in your browser using JavaScript. Your PDF is read from disk into memory, processed locally, and written back out as a download. Nothing is transmitted, stored or logged, which means contracts, invoices, medical records and identity documents stay on your device. The page also keeps working if you go offline after loading it.
Not in Lossless, Balanced or Strong mode. Those modes only re-encode embedded images and leave text, fonts and vector graphics byte-for-byte intact, so the document stays selectable, searchable and accessible to screen readers. Only Maximum mode rasterises pages, which converts text into pixels. Maximum is intended for scanned documents where the page is already an image.
It depends almost entirely on what is inside the file. PDFs full of photos or scanned pages commonly drop by 70 to 95 percent. Text-only PDFs exported from Word or LaTeX are already efficient and may only shrink by 2 to 15 percent in Lossless mode, because there are no images to reclaim space from. If a pass would make the file larger, the original is kept instead.
Use Balanced for almost everything - it keeps text intact and gives a large reduction on image-heavy files. Use Strong when you need to fit an email attachment limit and image fidelity is secondary. Use Lossless when the file must not change visually at all, for example for print or legal submission. Use Maximum only for scans, or when nothing else gets you under a hard size limit.
Yes. Enter a target such as 2MB or 500KB in Advanced options. The tool runs several compression passes and binary-searches the image quality setting to land as close to that ceiling as it can while staying under it. If the target cannot be reached even at the lowest quality, the closest result is returned and the file is flagged as target not reachable.
Yes. Drop up to 30 PDFs in at a time. They are processed one after another so the browser stays responsive, each with its own progress bar and before and after size. When they finish you can download them individually, merge them into a single PDF, or take them all as one ZIP archive.
There is no server limit because there is no server. The practical ceiling is your device memory, since the file and its decoded images have to fit in the browser tab. Files up to roughly 100 MB are comfortable on a desktop; on a phone, expect trouble well before that. If a large file fails, compress a page range at a time.
No. Encrypted PDFs cannot be parsed without the password, and the tool will report that the file is password protected. Remove the password in your PDF reader first, then compress the unlocked copy.
Almost certainly because it contains little or no image data, or its images are already well compressed. A PDF that is mostly text and vector graphics has no slack to reclaim. The other common cause is images in formats the browser cannot safely re-encode, such as CMYK JPEGs, JPEG 2000, JBIG2 or CCITT fax scans - those are deliberately skipped rather than risk corrupting the page.
Yes, usually by a further 10 to 30 percent on top of image compression, because colour information is discarded before the images are re-encoded. It is a good option for scanned text, receipts and forms where the colour carries no meaning. It is irreversible in the output file, so keep your original.
Optionally, and it is on by default. Removing metadata clears the title, author, subject, keywords, creator and producer fields along with the embedded XMP metadata block. That strips the name of whoever created the document and the software they used, which is worth doing before sharing a file publicly.
Yes to both. The page registers a service worker that stores the tool and its compression engine on your device, so it opens and compresses with no network connection at all. Your browser may also offer to install it as an app, which gives it its own window and launcher entry. Since nothing was ever being uploaded, offline mode is not a reduced version of the tool - it is the same tool.
Yes. It is completely free, requires no sign-up or email address, adds no watermarks, and has no daily limit on how many files you compress. Because the work happens on your own device there is no server cost to pass on.
