Most people arrive here because something refused their file. A portal capped at 5 MB, an email that bounced, a form that would not accept the upload. What actually makes a PDF that large is almost always photographs — a scanned page, a product shot, a site survey with twelve images across eight pages. Those are the megabytes.
The words themselves are described as instructions rather than pixels, so they cost almost nothing and lose nothing when the file shrinks. That is why a compressed document here stays selectable, searchable, and correct: the text was never the thing being compressed. The photographs are re-encoded at the amount you choose; scans and flat artwork are recompressed losslessly, and transparency is preserved.
You pick from three named amounts — or a size limit to get under. Gentle keeps the most quality, for something going to a printer where softening a photograph is not acceptable. Balanced is the one almost everyone wants. Smallest gives up the most, for a document that will only ever be read on a screen. There is no percentage to guess at, because the tool does not guess: it measures the images and shows an estimate first, then re-compresses the file and tells you the real size it reached, measured against the original.
Sometimes the honest answer is that there is nothing to do. A contract, a letter, a set of terms — text all the way through — is already close to as small as a PDF gets, and so is a document whose images are already efficient. Other tools hand you back a file 2% lighter and call it a success. This one tells you the truth and leaves your document alone, then points you at what might actually help: splitting it, or keeping only the pages you need.
Everything happens in this browser. There is no queue, no upload, no account, and no watermark on what comes out. If the network is off, it still works.