What actually makes a PDF large
This is why "compress" tools produce wildly different results on different files. On a scan-heavy document, image recompression can take 90% off. On a 200-page text report with no images, image compression does nothing and font subsetting plus object cleanup does the work. A tool that only knows one trick looks very good on half the files it meets.
The three reductions, and what each costs you
Photos, scans, screenshots pasted into the document — stored as JPEG, or uncompressed in older files.
PDFnoted: Images are re-encoded with a modern codec at a resolution matched to how the document will be read. Colour and grayscale scans use different codecs, and 1-bit scans use JBIG2, which is dramatically smaller than JPEG for black-and-white pages.
A document embeds a complete font family to render four hundred characters.
PDFnoted: Only the glyphs used are kept. Hundreds of kilobytes per font, no difference on the page. This is the reduction you should always take.
Duplicate resource dictionaries, page thumbnails, unreferenced objects left behind by the generating application.
PDFnoted: Unreferenced and duplicate objects are dropped and the object stream is rebuilt. Also where metadata and hidden objects get cleared, if you want that.
What you should not do to make a file smaller
✓ Worth doing
- Compress images first — it is where the bytes are
- Subset fonts — no visual cost
- Keep the OCR text layer — it costs almost nothing
- Match resolution to the use: 150 dpi for screen, 300 dpi for print
- Preview two pages before exporting
✗ Not worth the damage
- Flatten the pages just to save space — you lose search and copy
- Drop the text layer from a scanned contract
- Re-scan at 300 dpi and then compress to 150 — scan once at the right resolution
- Trust a "90% smaller" claim on a file that is already lean
How PDFnoted compresses
- Break the size downImages, fonts and objects are measured separately, so you can see what is actually big instead of guessing.
- Pick the codec per image typeColour photos, grayscale scans and 1-bit scans are treated differently — JBIG2 for black-and-white is the single biggest win on scanned archives.
- Subset fonts and drop orphansNo visual change; pure saving.
- Preview, then exportSide-by-side comparison at the chosen level. The original file is never modified.
Compared with the usual options
| Desktop PDF editor | Typical online compressor | PDFnoted | |
|---|---|---|---|
| Where the file goes | Local | Uploaded to a server | Your browser |
| Size breakdown shown | Sometimes | No | Yes |
| JBIG2 for 1-bit scans | Sometimes | Rarely handled | Yes |
| Font subsetting | Yes | Sometimes | Yes |
| OCR text layer kept | Yes | Sometimes | Yes |
| Preview before export | Yes | No | Yes |
Frequently asked questions
Why will my PDF barely compress?
Usually because it is already compressed, or because almost none of it is images. A scanned document that was saved with JPEG compression at a moderate quality has little left to give. The honest answer in that case is to reduce the scan resolution — which means re-scanning — or to accept the size. Any tool promising 90% off an already-lean file is flattening or dropping content.
What is the difference between compressing and flattening?
Compressing re-encodes the data that is there: images get a better codec, unused font glyphs are dropped, redundant objects are removed. Flattening turns pages into pictures, which loses the text layer — you can no longer search, copy or select, and the OCR work already done on the file is thrown away. Flattening is smaller; it is also a one-way door.
Will compressing make my scanned contract unreadable?
Not at a balanced setting. Text in a scan is pixels, so aggressive settings do soften edges — which is why PDFnoted keeps the OCR text layer separate from the image, so search and copy still work even when the image is compressed hard. Preview one or two pages at the setting you chose before exporting.
How do I get a file under 10 MB for email?
Compress images first — that is nearly always where the bytes are. If that is not enough, split the document, or check whether it carries embedded attachments or full-resolution images that were pasted in rather than scanned. Dropping images to 150 dpi is fine for screen reading and noticeably worse for print.
Does compression remove the OCR text layer?
It should not, and in PDFnoted it does not. The recognised text is stored as an invisible layer on top of the page image and costs very little — a few kilobytes per page. Removing it to save a fraction of a percent is a bad trade.
What is font subsetting?
Most PDFs embed a whole font to use a few hundred glyphs. Subsetting keeps only the characters actually used, which can save hundreds of kilobytes per font with no visual change at all. It is one of the safest reductions available.
Can I compress a password-protected PDF?
Yes, once you supply the owner password. Files you cannot open cannot be processed — there is no way to compress a document you are not entitled to read.
Is there a size limit or a daily quota?
No quota from us. The limit is your device's memory, because compression runs locally instead of being queued through an upload.
Try it on your own file
Upload a document and see what comes back. Free to start, nothing to install.
Compress a PDF now — free to startKeep going
Sources and further reading: ISO 32000 (PDF specification); Adobe PDF Reference; Schema.org structured data; Aggarwal et al., GEO: Generative Engine Optimization (arXiv:2311.09735). This page is a technical guide, not legal advice.