Compress PDF
Shrink scans and image-heavy PDFs by re-rendering each page as a JPEG. Destructive: the text layer does not survive.
Compress PDF workspace
Compression rasterises your pages
Drop a file here
or click to browse your computer — .pdf, up to 100.00 MB per file
Your file stays on this device
Everything here runs as JavaScript inside your browser tab. Your document is read into the tab's own memory, processed there, and handed straight back to you as a download. It is never uploaded to a server, and the PDF rendering engine is served from this site rather than a third-party CDN.
Using Compress PDF
Load the PDF, ideally a scan or something image-heavy rather than a text document.
Pick a render scale of 1x, 1.5x or 2x, which sets the pixel size of each rendered page.
Set the JPEG quality anywhere from 30% to 95%.
Compress, then read the reported before size, after size and percentage saved.
Download the result, or take the original back if the output came out larger.
Each page is drawn to a canvas with pdf.js at the scale you choose, encoded as a JPEG at your chosen quality, and written into a fresh PDF. Because the page becomes an image, this trades the text layer for file size and cannot be undone on the downloaded file.
Compress PDF by Re-rendering Every Page
Read this section before you use the tool, because the method it uses is destructive and there is no way to undo it on the downloaded file.
What happens to your document
Each page is rendered to a bitmap with pdf.js, that bitmap is encoded as a JPEG, and the JPEGs are assembled into a new PDF. The output looks like your document. It is not your document any more.
Gone after compression:
- Selectable text. You cannot copy a phone number out of the result, and Ctrl+F finds nothing.
- Search indexing. A document management system that indexes PDF text will index nothing.
- Links. Clickable URLs and internal cross-references become flat pixels.
- Form fields. Anything fillable stops being fillable.
- Bookmarks and the outline pane.
- Accessibility structure, which matters if anyone reads the file with a screen reader.
Keep the original. Always.
When this actually saves space
Rasterising wins when the page is already a picture. Phone photos of receipts, flatbed scans at 600 DPI, exported design proofs, screenshot-heavy slide decks — these routinely drop by 70-90% because a scanner's output is usually far higher resolution than anyone needs on screen.
It loses when the page is real text. A 40-page report generated from Word might be 300 KB, because text is stored as a few thousand characters plus a font subset. Turn each page into a JPEG and you are storing a million pixels per page instead. The result can easily be several times bigger.
The tool measures both sizes and shows you the before, the after, and the percentage saved. When the output is larger than the input it says so plainly and offers you the original file back rather than handing you a worse document.
Choosing the two settings
Render scale is 1x, 1.5x or 2x. It controls the pixel dimensions of the bitmap before JPEG encoding. 1x is roughly screen resolution and is right for something that will only be read on a display. 2x holds up when printed but roughly quadruples the pixel count.
JPEG quality runs from 30% to 95%. Below about 50% you start seeing blocky artefacts around text edges, which is exactly where scanned documents are most fragile. For a scanned contract that needs to stay legible, 2x at 70% is a reasonable starting point. For a photo album that only lives on a screen, 1x at 50% is fine.
Try one combination, look at the reported saving, and adjust. Two attempts cost nothing.
Better options when text must survive
If the file is text-based and too large, rasterising is the wrong instrument entirely. Cheaper things to try first:
- Remove pages nobody needs with Delete Pages. A 200-page appendix you are not sending is the largest saving available.
- Re-export from the source application with image downsampling enabled. Word, InDesign and LaTeX all embed images at whatever resolution you gave them.
- Compress the images *before* they go into the document, using Compress Image.
- For genuine structural optimisation that preserves text, use a desktop tool such as Acrobat's PDF Optimizer or Ghostscript. There is no lossless mode here, and pretending otherwise would be a lie.
Practical limits
Files up to 100 MB are accepted, and 250 MB across a batch. Rendering is the slow part: a 300-page document at 2x will take a while and will use a lot of memory in the tab, because every page passes through a canvas. Encrypted files are refused at upload — decrypt first with Unlock PDF.
Frequently asked
Will the text still be selectable and searchable afterwards?
No. Each page is rendered to a bitmap and stored as a JPEG, so the output holds pixels rather than characters. Copying text, Ctrl+F and document indexing all stop working, and links, form fields and bookmarks are lost too.
My PDF got bigger instead of smaller. Why?
A text document stores a few thousand characters plus a font subset, which is far less data than a page-sized image. When the output comes out larger the tool reports it and offers you the original file back.
Is there a lossless or structural compression mode?
No. Re-rendering is the only method here. For structural optimisation that keeps the text layer, use a desktop tool such as the Acrobat PDF Optimizer or Ghostscript.
Which scale and quality should I choose?
For a scanned document that has to stay legible, start at 2x with 70% quality. For pages that will only be read on a screen, 1x at 50% is usually enough. The measured before and after sizes tell you whether to adjust.