How the tool reaches an exact size
Compression here is a search, not a slider. The tool renders each page, encodes the pages as JPEG images at a series of quality levels, and keeps the highest quality whose rebuilt PDF still fits under your cap. That is why the result lands just under the target instead of far below it.
When even the lowest sensible quality overshoots the cap, the pages are rendered at a lower resolution and the search runs again — and the result line tells you the resolution it used, in dpi (dots per inch). A sharper page at lower resolution almost always beats a full-resolution page ground into artifacts, which is why scaling is the fallback rather than pushing quality to zero.
For scans of text pages there is a second lever: black-and-white mode. A grayscale page needs far fewer bytes than a colour one, so switching it on often keeps more detail at the same target size.
What re-encoding does to a scanned document
A scan is already a picture of a page — there is no text inside it to lose. Re-encoding that picture at a lower quality trades fine grain (paper texture, print edges) for file size, and at 150 dpi the words stay perfectly readable. That is the same trade every PDF compressor makes, including the ones that upload your file.
A PDF with real text in it is a different case: its letters are stored as text, which is why a five-page letter is often smaller than one photo. Rebuilding such a page as an image makes the text part of the picture — readable, but no longer selectable or searchable. If your document is text-only and still under the cap after that, the tool leaves it completely unchanged.
Why portals ask for 100 KB in the first place
Government portals, job applications and university forms cap the file size because storage and virus-scanning cost scale with bytes — the tightest ask for 100 KB, others draw the line at a few megabytes. The cap does not care how many pages you have, which is why a ten-page scan can become a puzzle: the portal wants fewer bytes, not fewer pages. That is the moment a compressor that aims at an exact size is the right tool, and a preset like “medium quality” is not.
If it is a photo, not a document, that must fit under a cap, the image compressor does the same search without touching PDFs at all. To combine several PDFs first, use the PDF merger; to cut pages out first, the PDF splitter.
Why nothing is uploaded
The file people need under 100 KB is usually their least anonymous one: an ID, a payslip, a signed contract, a certificate for a visa application. On an upload-based compressor that file travels to a server you know nothing about, and the promise to delete it is a sentence on a page, not something you can check.
Here the work happens in your browser with standard web APIs. There is no upload, no queue and no file size limit beyond your device’s memory; the one hard cap is 50 pages per document. You can watch the network tab while you use it: no request carries your document. More on the difference in why upload-based tools are a privacy problem.
Questions
How exact is the target size?
The result is always at or under the target, usually within a few percent of it. The tool keeps the highest quality that fits, so it does not land far below the cap on purpose.
What happens to my document?
Each page is re-rendered and re-encoded as an image, then the pages are rebuilt into a new PDF. Text on the pages stops being selectable text — it becomes part of the page image. For uploading to a portal that only checks the file size, that is exactly what you want.
Which documents does this work best on?
Scans and photos of pages — contracts, certificates, IDs, application forms. Those are already images inside the PDF, so re-encoding them loses almost nothing. A text-only PDF is usually small to begin with and rarely needs this.
What if the target cannot be reached?
With very many pages under a very small cap, no setting fits. The tool then scales the pages down, and if even that is not enough it says so and gives you the smallest file it could produce.
Can it read password-protected PDFs?
No. A PDF that needs a password to open cannot be read here, by design — there is no server that could take the password.
Is my document uploaded anywhere?
No. Rendering and re-encoding run in your browser. You can verify it in your network tab: no request carries your document.