Compress PDF — Free Online Tool on Toolpile
Every tool you need, one place
PDF, Image, AI, Dev, and Business tools — free, private, no signup.
Toolpile uses cookies for analytics and ads. The tools themselves keep your files in the browser — that doesn't change. Read the full policy
PDF, Image, AI, Dev, and Business tools — free, private, no signup.
About Compress PDF
PDF compression is the most-googled PDF operation after "PDF to Word" — and the one most online tools quietly do badly. The reason: a PDF isn't really a file, it's a container of fonts, images, and vector content with very different compression ceilings. A 50 MB scanned PDF often shrinks 80%; a 5 MB text-only PDF rarely shrinks more than 5%. Knowing why is the difference between "the tool didn't work" and "that file was already compressed".
Open any 10 MB PDF and 95% of the bytes are usually one of three things: embedded images (scanned pages, photos in reports, screenshots), embedded font subsets (every PDF carries the glyphs for every font it uses, so the file renders identically anywhere), and uncompressed content streams (the page-drawing instructions). Reducing the file size means working on those three layers — there's no magic "shrink" that operates on the PDF as a whole.
Scanned PDFs are the dramatic case. A 20-page scan at 300 DPI is roughly 20 × 8.5 × 11 × 300² × 3 bytes uncompressed = ~470 MB of pixel data, stored in the PDF as JPEG-compressed images that come out around 30-80 MB depending on quality. Re-encoding those JPEGs at 150 DPI and quality 75 — perfectly readable for screen viewing — reliably hits 80-90% reduction. The original is never seen again, and that's fine if you don't need archival fidelity.
Text-heavy PDFs are the disappointing case. A 100-page contract from Word is maybe 2 MB; the actual text is a few hundred KB and the rest is font subsets. You can subset the fonts harder (drop unused glyphs), strip metadata, and recompress the streams with Flate — total saving usually 5-15%. Anyone promising 90% off a text PDF is either silently downsampling embedded images or showing you a broken file.
PDF supports two relevant image filters: DCTDecode (JPEG-style lossy, 8×8 block discrete cosine transform) and FlateDecode + predictors (PNG-style lossless). Photos and scans should go through DCTDecode at quality 70-80 and downsample to 150 DPI for screen, 300 DPI for print. Diagrams, screenshots, charts, and anything with sharp edges or text-as-image should stay in lossless mode — JPEG produces visible ringing artefacts around text edges that look like a printer error, even at quality 90.
The cheap online compressors that produce "smaller PDFs that look terrible" are the ones that JPEG-compress everything indiscriminately. Better compressors detect image type per-image — high-frequency content (photos) gets DCT, low-frequency or text-heavy content stays lossless. This tool tries that distinction, but the source matters: if your scanner saved every page as one big JPEG (most do), the recompressor can't undo that — it can only re-JPEG it.
If your output looks blocky around text after compression, the source had text rendered as part of a JPEG image. Re-OCR'ing the document (extract text, drop image, render fresh) is the only fix that gets sharp text back, and that's a different tool — see PDF to Text or PDF to Word as a starting point.
Every PDF embeds the fonts it uses so it renders identically everywhere. Fonts come in two embed modes: full embed (the entire font file, often 200-500 KB per typeface) or subset (only the glyphs actually used, typically 20-50 KB). A PDF generated from Word with default settings often full-embeds fonts; a PDF generated from a modern web tool (Puppeteer, headless Chrome, pdf-lib) usually subsets automatically. If you compress a Word-exported PDF and see a 30-50% reduction, font subsetting is doing the work, not image recompression.
Already-subsetted PDFs (LaTeX output, modern web exports, files already through a compressor) have very little room left here. This is why running the same PDF through compressors twice produces the same output — the second pass has nothing to do.
One trap: if your PDF uses lots of typefaces (12 different fonts in a marketing brochure), subsetting still leaves 12 small subset blobs. Consolidating to 2-3 fonts at design time saves more than any post-processing compressor can.
Why didn't my PDF get smaller?
Almost always one of three reasons: (1) the PDF is text-only and already efficiently encoded — there's nothing to compress, (2) the source was already through a compressor (modern Office, Google Docs, and online tools all output near-optimal files), or (3) the embedded images are already heavily JPEG-compressed and re-encoding them further would visibly degrade quality. Some files are simply at their floor.
Will compression make my PDF unsearchable or break selection?
No, as long as the text was real text in the source. Compression operates on streams (Flate, DCT) and image data; the text-extraction layer (a parallel content stream) is preserved. Where it breaks: if the source was a scan with OCR'd text on top, aggressive image downsampling can produce a mismatch between the visible image (now lower-res) and the OCR text layer (still aligned to the original resolution). The text still extracts; selection highlighting may look slightly off.
How does this compare to Adobe Acrobat's compression?
Acrobat Pro has access to per-image type detection, JPEG2000 (smaller than JPEG at the same quality), and CCITT Group 4 (specialist black-and-white fax compression that beats anything else for pure-text scans). Browser-based compressors like this one use JPEG and Flate — universal, but ~10-20% larger output than Acrobat at the same visible quality. The trade is privacy (Acrobat Online uploads your file; this tool doesn't) and cost (Acrobat Pro is paid).
Will it preserve digital signatures?
No — and no compressor can. A digital signature signs a specific byte range of the PDF; any modification, including recompressing an image, invalidates the signature by design. If you need a signed PDF smaller, compress before signing. If the file is already signed and must stay signed, you cannot also compress it; pick one.
Is there a maximum file size?
Browser memory limits practical compression to ~500 MB. Past that, Chrome's per-tab memory budget pushes back and the tab can crash mid-process. For multi-GB scans, a desktop tool (Acrobat, Preview on macOS, PDF24 desktop) is more reliable.