File compression
How to Compress a PDF: The Trade-offs and Common Mistakes
Attachment too large, uploads timing out, a submission system rejecting your file — all of these point at the same action: compressing a PDF. It is worth understanding what compression actually does before you do, because that determines how much quality you give up.
What compression actually targets
A PDF is a container that holds three kinds of content: text and vector graphics (stored as instructions, very small), raster images (scans, photos), and embedded fonts (usually stored whole). "Compression" is a trade-off between them.
- Lowering image resolution — resampling a 300dpi scan to 150dpi roughly halves the size; small text starts to soften.
- Re-encoding images as JPEG — replacing lossless PNG with 85% quality JPG typically saves 60–80%.
- Removing unused objects — dropping unreferenced images, hidden layers, redundant metadata. This step is lossless, but most files have little of this to remove.
- Subsetting embedded fonts — real but modest gains, and it can affect how the file renders on someone else's machine.
The point: the first two are lossy, the third is not. If a file shrank by 80%, it almost certainly means its images were re-encoded, and quality was paid for.
Scans and text documents need different treatment
Scanned / image-based PDFs (contracts, invoices, book pages) carry nearly all their weight in embedded images. There is a lot of room to work with, which makes them the best candidates. UniPagePDF re-renders each page as JPEG at the chosen resolution, typically saving 40–70%.
Text-based PDFs (exported from Word, generated reports) are already compact, with most of their weight in fonts. The only lever left is resolution, so the gains are small and the visual cost is real. For those files, keeping the original is often the better answer.
Choosing a level
- Strong — for meeting a hard upload limit, such as a 2MB submission cap. Text visibly softens; acceptable for on-screen review only.
- Balanced — the recommended default. Usually indistinguishable by eye while still delivering a solid reduction.
- Light — for printing or long-term archiving, where legibility comes first.
A practical approach: look at the target system's size limit and work backwards. If the cap is 50MB and your file is 12MB, you did not need to compress anything.
Common mistakes
Mistake one: assuming stripping metadata does the heavy lifting. It helps a little — usually a few percent. The real weight is in images; removing the author and title barely moves the number.
Mistake two: smaller is always better. If a contract compresses until the signature date is unreadable, the bytes you saved bought nothing. Use Light or Balanced for contracts, medical records, and certificates.
Mistake three: compressing before OCR. Compression directly lowers recognition rates. OCR first and compress afterwards, or stick to the Light level.
The privacy upside
UniPagePDF compresses in your browser, so nothing is uploaded. Compressing a contract, a financial report, or a medical record creates no additional copy anywhere — the file exists only on your device and disappears when you close the tab.