Compress PDF Online — Reduce File Size Free
Part of PDF tools: See all PDF tools.
What is PDF Compress?
This free PDF compressor shrinks a PDF so it fits under common email and portal size caps. You upload the file over HTTPS. At the low setting the tool only rebuilds the file: unused objects are removed and streams are deflated. Medium and high also re-encode embedded raster images as JPEG (quality about 50 on medium, about 25 on high; high also halves very large images). Text and vector drawing stay as vectors. There is no guaranteed percentage of savings — image-heavy scans shrink a lot, text-only PDFs often barely move. The compressed file is returned as a download and the upload is discarded. No account, no watermark.
How to use PDF Compress
- Open https://gotoolsonline.com/tools/pdf-compress and upload a PDF (or switch to Multiple PDFs for a ZIP of results).
- Choose Low if you want a tidy rebuild without touching images, Medium for a balance of size and quality, or High when a portal still rejects the file.
- Click Compress & Download. Compare the original and compressed sizes on the page, then open the download and check a text page and an image page at full zoom.
- If the file is still over the limit, try High, or split the PDF and compress each part. Keep your master copy; send the compressed one.
Why use this tool?
Email caps (often about 25 MB) and government or college portals (often 5–10 MB) reject oversized PDFs every day. Desktop Acrobat is more than most people need for a one-off shrink. This page is for that job: clear low / medium / high behaviour, honest expectations when the file is mostly text, and a download you can verify before you send it.
Try-and-check for size limits: Start with a moderate compression setting, download the result, then verify the output size. If it's still too large for your upload form, switch to High (Medium is the default). If High is still too large, send fewer pages with Split PDF: a second pass at the same level saved under 1% in our tests.
For best readability, keep an eye on image-heavy pages - text-heavy PDFs usually compress more cleanly.
The anatomy of a bloated PDF
Before you can compress a PDF intelligently, it helps to know what is actually taking up space inside the file. A PDF is a container. Inside it are object streams — little blobs of data that describe pages, fonts, embedded images, form fields, annotations, and sometimes chunks of JavaScript or multimedia. For most documents people need to shrink, the culprit is one specific thing: embedded raster images. A scanned contract at 300 DPI with full-colour pages can easily run 800 kilobytes per page, and a 40-page scan will clear 30 MB before anything else contributes. A text-only thesis with a handful of vector diagrams, by contrast, might be a single megabyte for hundreds of pages.
The second largest offender is font subsetting failure. Some PDF generators embed the entire font file — every glyph of every weight — even though the document uses 60 characters of one weight. A single TTF can weigh 400 kilobytes. Multiply that across a family, and you have added two or three megabytes of invisible payload. The third offender is metadata and object-stream bloat: incremental edits in Acrobat or mobile scanner apps often append new versions of objects without removing the originals, so the file literally contains every revision of every page stacked on top of each other.
How this compressor actually works
Three levels map to real behaviour. Low rebuilds the PDF: unused and duplicate objects are garbage-collected, content streams are cleaned, and remaining streams are deflated. Images are left as they are. Medium does the same rebuild after re-encoding most embedded raster images as JPEG at quality about 50, with a correct /DCTDecode filter and matching colour space. High uses JPEG quality about 25 and halves images whose longest side is over about 2,000 pixels before encoding. Tiny icons, 1-bit masks, and specialised codecs (JPEG2000, JBIG2, CCITT) are skipped. Soft masks that sit beside an image are left in place; colour-key masks that no longer match the new JPEG are dropped.
Text and vector drawing are not rasterised. Zoom in on a born-digital letter and the outlines are still vectors. Softness appears on photos and scans first. There is no separate font-subsetting pass and no automatic CCITT conversion for bitonal scans — those claims belonged to older marketing copy, not this code path.
Realistic compression expectations by document type
Scanned colour documents are where compression works best. A 40-page colour scan at 300 DPI usually lands between 30 MB and 60 MB uncompressed. After a medium-strength pass it typically sits between 4 MB and 10 MB — well within the 25 MB Gmail attachment cap. Black-and-white text scans, surprisingly, do not shrink as dramatically in percent terms because they are already efficiently encoded — but they often drop from 10 MB to 2 MB, which is still a big practical win.
Documents that were exported directly from Word, Google Docs, or LaTeX — born-digital files that were never a physical page — compress much less. The images inside are usually PNG diagrams or vector graphics, and the text is already an efficient vector stream. Expect 10–25% reduction at most, sometimes nothing. If you see "compressed PDF is only 4% smaller", that is usually why: the source was already compact.
PDFs that contain mixed content — a native export with a few scanned receipts appended at the end — often surprise people. The 2 MB born-digital portion stays roughly the same; the 20 MB scanned appendix collapses to 3 MB. The overall file drops from 22 MB to 5 MB, which reads as "77% reduction". Different pages, different outcomes, same pass.
When not to compress
If the PDF is going to be run through OCR later, or used as archival evidence in a legal matter, do not compress it. OCR accuracy drops measurably on reprocessed JPEG images because JPEG's chroma subsampling blurs letter edges slightly. For scanned court documents, government submissions with explicit DPI requirements, or anything destined for long-term preservation, keep the source at full resolution and compress only the copy you are emailing today.
Forms with fillable fields are generally safe — field objects are unchanged by image recompression. But digitally signed PDFs are a special case: a cryptographic signature covers the exact byte sequence of the file when it was signed. Compressing a signed PDF always invalidates the signature. If you need a smaller signed PDF, sign a smaller source first; do not sign then compress.
Try-and-check workflow
The most reliable way to use any PDF compressor is iterative. Start with the default or medium setting, download the result, open it, and scroll through a few of the image-heavy pages at 100% zoom. If the result looks acceptable, ship it. If it is too large for your target upload slot (a 10 MB HR portal, a 5 MB visa application), compress the original again on High; a second pass on the download saves very little, so send fewer pages if High is still too big. If the result has visible blocking or smearing on photographs, step the quality back up one level.
Most users never need to touch the strong setting — medium handles 90% of everyday compression jobs. The strong setting is there for the rare cases where a 60 MB scan has to fit into a government portal that refuses anything above 3 MB. At that point, you are trading quality for acceptance, and the tool makes the trade visible: you can see the compressed output before you submit it.
Privacy notes specific to PDF uploads
PDFs are processed in server memory. The original file exists in RAM for the duration of the compression request, which for a 50 MB input is typically 3–8 seconds, and is garbage-collected immediately after the compressed bytes ship to your browser. The file is not written to disk, not cached, not logged. The server logs the request (method, path, response size, response code) but not the file contents or the filename. If the request fails, the partial data in memory is discarded without a trace. This is the same handling every file tool on this site uses.
If you are compressing something sensitive — a medical record, a tax return, a passport scan — remember that compression is about size, not privacy. Document metadata (author, editing application, creation date) is tiny compared to the image payload and can survive compression, so open the compressed file's document properties in your PDF reader before you send it. If those fields reveal more than you want, clear them in the program that created the PDF and export again; that is a separate, intentional step.
What to do if the output is still too large
Occasionally a PDF refuses to shrink below a target threshold no matter how aggressively you compress. There are three likely reasons, and each has a different fix. First, the document may be almost entirely embedded scans at extreme resolution — 600 DPI archival scans, in particular, sometimes resist further reduction because their quality-to-size curve flattens early. In that case, run split PDF to break the file into halves, compress each half, and distribute as a two-part package. Second, the document may contain one or two rogue full-page images at resolutions that are genuinely wasteful (a 40 MB PNG logo on the cover page is not uncommon). Opening the PDF in a viewer and checking which pages are heavy helps identify the culprit. Third, the file may have absorbed many rounds of incremental edits, each leaving behind superseded object streams; a single compression pass should clear those, but if the pass ran on an extremely damaged file, re-exporting from the original source is sometimes the only clean fix.
A practical upper bound: a visually lossless compression of a 40-page colour scan generally cannot go below about 1 MB without quality degradation becoming obvious at 100% zoom. If your target is below 1 MB for that kind of content, you will need to accept either visible quality loss or page splitting. The compressor's strong setting pushes into visibly-compressed territory on purpose; use it when acceptance matters more than visual fidelity.
Target sizes worth aiming for
Under 25 MB: most consumer email caps for a single attachment. Under 10 MB: common corporate and government forms. Under 5 MB: picky LMS and visa portals. Under 2 MB: aggressive mobile email or multi-file messages. Hit the next larger tier first; only use High if the recipient’s rules demand it.
What a useful compress page teaches
A thin page only repeats “compress PDF online free.” A useful page teaches the order of levers (low → medium → high → split), when size will not move (already-lean text PDFs), and how to verify the download before you email it. That is the bar this page is written to clear.
Frequently asked questions
- Is PDF compress free?
- Yes. No signup and no watermark on the download.
- Is my PDF safe?
- It is uploaded over HTTPS, processed for this request, returned as a download, and discarded afterward. We do not keep a permanent archive. See the Privacy Policy.
- What do Low, Medium and High do?
- Low rebuilds the PDF with unused objects removed and streams compressed; images are not re-encoded. Medium re-encodes raster images as JPEG at quality about 50. High uses JPEG quality about 25 and halves images whose longest side is over about 2,000 pixels. Both medium and high still run the low rebuild afterward.
- How much smaller will my PDF get?
- There is no fixed percentage. Scanned or photo-heavy PDFs often shrink a lot. Born-digital text PDFs with few images may only drop a little, or stay almost the same. Always check the size shown after compression.
- Will the text stay sharp?
- Text and vector drawing are not turned into JPEG. Softness appears on photos and scans first. Spot-check a text page and an image page at 100% zoom before you send the file.
- What if I still exceed Gmail or a portal limit?
- Try High, or use Split PDF and compress each part. For photo-only attachments, compress the images first with Image Compressor, then rebuild or attach them separately.
Step-by-step guides
Also try
Related tools that work well with this one: