Keep existing image pixels
The default mode compacts document data without resampling or re-encoding images. Text and vector pages are not converted into screenshots. Existing image pixels stay unchanged.
Compress PDF
Make a smaller PDF locally, then compare the actual file sizes before downloading.
No file selected
Keep image pixels unchanged. Already optimized files may not get smaller.
One PDF · 30 MiB · up to 60 pages · no forms or signatures
Choose one PDF to start.
- Original
- Download
In your browserNo uploadsNo account
How to compress a PDF
- Choose your PDFChoose one unprotected PDF from your device.
- Choose a compression modeKeep image pixels unchanged, or explicitly choose lossy JPEG optimization. Select Compress PDF.
- Compare and downloadCompare the original and download sizes, then download and inspect the result. If the candidate is not smaller, the download remains your unchanged original.
Reduce PDF size and control image changes
Optional JPEG optimization
The optional image mode also re-encodes eligible RGB JPEG images at 78% quality and reduces their longest edge to at most 2,000 pixels. Fine print, photos and gradients can lose detail. An image is replaced only if its new encoded data is smaller. Other image types are left unchanged.
Actual savings, not a target size
The result shows actual bytes and savings, not an estimated percentage. A PDF that is already optimized may not shrink. This tool does not target an exact size such as 100 KB or 1 MB, remove pages or promise a fixed reduction.
Support and limits
- One unprotected PDF, at most 30 MiB and 60 pages. Forms, signed or certified documents and encrypted PDFs are rejected rather than rewritten.
- JPEG optimization is limited to simple 8-bit RGB JPEGs without masks or custom decoding. Each decoded image is limited to 12 million pixels, with 60 million decoded pixels in a run. Unsupported or oversized images are skipped.
- Compressed images in other formats, scanned text recognition, font subsetting, repair of broken PDFs and guaranteed PDF/A or accessibility conformance are outside this tool's scope.
- Processing stops above 15,000 document objects, 45 seconds of worker processing or the 60-second overall guard. Device memory and browser support can impose lower practical limits.
- Review page count, small text, images and links in your own result. Tested fixtures do not establish compatibility with every PDF. Do not use this tool to preserve digital signatures or an archival certification.
Test files and results
On October 25, 2026, we selected these original English PDFs in this tool, downloaded the results and visually checked all six output pages. Each run kept both source pages.
| Input and mode | Actual download and bytes | What changed |
|---|---|---|
| Text and links PDF Keep image pixels unchanged | Actual text PDF result 11,269 → 2,924 bytes | All 18 table rows on each page, end markers, text and two link targets retained. Both page renders matched the original at the comparison resolution. The source uses uncompressed page data. |
| Image-heavy PDF Keep image pixels unchanged | Actual unchanged-image result 2,379,982 → 2,379,327 bytes | Only 655 bytes saved. The shared JPEG stayed at 1,800 × 2,400 pixels with identical image data. Both page renders matched the original. |
| The same image-heavy PDF Optimize eligible JPEGs · lossy | Actual lossy-image result 2,379,982 → 354,201 bytes | The JPEG became 1,500 × 2,000 pixels. All nine check labels, amounts, border and bottom marker remained visible on both pages, but fine background texture changed. The native text captions remained text; the image was not OCRed. |
Running the 2,924-byte text result through the default mode again saved zero bytes. The downloaded original was byte-identical. A separate 390-pixel browser-width run completed file selection, keyboard conversion and download; its text PDF matched the desktop output. This was not a physical-phone test.
Encrypted, fillable-form, malformed, 61-page and over-30-MiB inputs were rejected without an output download. The successful 60-page limit, object-count limit, image-pixel budgets and processing-time limits were not driven to their boundaries.
Download the page-by-page check record and source notes. Savings are specific to these files; no general reduction percentage is promised.
Local processing and privacy
Your PDF and the optimized file stay in this browser. File contents and filenames are not sent to a conversion server. Conversion components load from this site. Page delivery and analytics are separate requests; document contents and conversion results are not included in analytics events. See the privacy policy for page-visit data.
Close the page when finished. Files already downloaded remain on your device.
There is no daily conversion quota or added watermark.
Frequently asked questions
Why did my PDF stay the same size?
Its existing compression may already be efficient, or its images may not qualify for optional JPEG optimization. If the new candidate is not smaller, this page offers your original file and reports zero bytes saved.
Can I keep a digital signature or a fillable form?
No. PDFs containing forms, signature fields or certification restrictions are rejected. Keep the original signed or fillable document.
Page updated: