Compress a PDF — without uploading it anywhere

What actually happens to your file

The tool reads your PDF's internal structure, finds every image inside it, and re-encodes each one at a lower resolution and a tuned quality setting, using a proper JPEG encoder (mozjpeg, compiled to WebAssembly and served from this site). Text, vector drawings and layout are untouched — only the images are rebuilt, and the file is rewritten around them.

The legibility governor

Every re-encoded image is scored against the original before it is accepted. If the letters in a scan would come out thinner than the governor's floor — if the result would be visibly worse at reading size — the tool backs off the resampling for that image instead of shipping grey soup. This is a measured check on every image, not a quality slider you have to guess with.

Three honest outcomes

Depending on what your file actually is, you will get one of three results, each stated plainly. If the file has real images, they are re-encoded and you see the before/after. If the file has no images — a Word export, say — there is no quality to trade away, so the tool restructures it losslessly and says exactly that. And if every image in your file is already stored as well as it can be (a fax, for instance, whose bilevel encoding is already tight), the tool keeps your original and tells you so. "Kept as-is" is a correct result, not a failure — a smaller number would have cost you quality that the file didn't have to spare.

For the same reason, a quality picker only appears when the choice can change the answer. When measurement shows every preset would produce byte-identical output, offering the picker would be theatre, so it isn't offered.

What it costs

Careful compression is slower than careless compression — expect a few seconds per large image, with per-image progress and a Cancel button that actually cancels and leaves your original untouched.

Questions people actually ask

Why is my PDF so big in the first place?

Almost always: it contains photographs. A phone-scanned page is a multi-megapixel photo, and a ten-page scan is ten of them. Re-encoding those images at reading resolution is where nearly all the savings live.

Why did I get "kept as-is" instead of a smaller file?

Because the tool measured every image in your file and found nothing it could improve without visible damage. Some files — faxes and bilevel scans especially — are already packed as tightly as they honestly get. You get your original back and a plain statement of why. That is the correct result.

Why is there no quality slider on my file?

The tool inspects your file first. If a choice of preset cannot change the output — for example, a text-only document with no images — no picker is shown, because a picker that cannot change the answer would be a decoration, not a control.

Will the text stay sharp?

Yes, by measurement rather than promise: a legibility governor scores every re-encoded image and backs off the resampling if letter strokes would drop below its thickness floor.

Is the file uploaded to be compressed?

No. The compression runs in your browser on your device. Watch the Network tab while it runs — no request leaves with your file.