Will my text still be selectable afterwards?
Yes, always. The text, the fonts and the vector drawings are copied into the new file untouched. Only the photographs are re-encoded, and only when re-encoding them actually saves something.
Compress a PDF, honestly
Almost all the weight in a large PDF is photographs, so those are what get re-encoded. The text, the fonts and the drawings are left exactly as they were — and if the result is not meaningfully smaller, you are offered nothing rather than a pointless second copy.
Measured against your original
—
Nothing compressed yet
What this tool does, exactly
- What it does
- Finds the JPEG images inside the document, re-encodes them smaller, puts them back, drops the metadata, and rewrites the file with object streams. Text, fonts and vector drawings are copied through untouched, so the result is still selectable, searchable and readable aloud.
- What it refuses to do
- Rasterise the pages. Turning every page into a picture is how most “compress PDF” tools get their impressive numbers, and it destroys selectable text, screen-reader access and print quality to do it. That is a different operation, and on this site it is called redaction.
- When you get nothing
- When there was nothing to win. If the document holds no images, or its images are already efficiently stored, or the saving is under one per cent, or the rebuild came out larger, no file is offered and the reason is named. Your original is already the smallest version of it.
- Why the numbers vary by device
- Browsers do not agree on what a quality setting means: the same photograph re-encoded at the same nominal quality came out 31% smaller in Chrome, 19% smaller in Firefox and 30% larger in Safari. So the tool encodes, measures, and steps the quality down until the result is genuinely smaller — rather than trusting a number.
About this tool
Why was I offered no file at all?
Because the compression was not worth taking. There are four ways that happens: the document holds no images to work on, its images are already efficiently stored, the saving came to less than one per cent, or the rebuilt file was larger than the original. In every case the result panel says which, and your original file is already the smallest version.
Why did I get less compression than a friend on another device?
Because the work is done by your browser's own image encoder, and they do not agree with each other. The tool compensates by encoding, measuring the result and stepping down until it is genuinely smaller — so everyone gets a smaller file, but not the same file.
Does it remove hidden data?
It removes the document's metadata: the title, author, producer, creation and modification dates, and the XMP packet that often carries the full editing history and the author's name. It does not remove annotations, form fields or attachments — that is what the redaction tool is for.
What do the three levels do?
They set how far photographs may be scaled down and how hard they are compressed. Light keeps photographs close to the original, balanced is a good default for email and uploads, and strong produces the smallest file with visible softening of photographs. None of them touches the text.