Re-save a PDF with object streams to trim its size, in your browser. Nothing is re-encoded, so the image and font data is kept exactly as it was. The file never leaves your device.
This tool takes PDF documents — pick one to select a file.
Trim a PDF's size by re-saving it with object streams, in the browser. Nothing is re-encoded, so the images and fonts inside are copied exactly as they were, and the file never leaves your device — the work happens in this tab. Up to 200 MB per document, and every page is kept.
The document is loaded and written out again, and that is the whole mechanism. The one real lever is packing the document's object definitions into object streams — which is a genuine saving on a PDF made of many small objects, and close to nothing on one whose weight is in its images. No image is decoded or resampled, there is no quality setting to lower, the fonts are copied rather than rebuilt, and the metadata is carried through unchanged. A tool that offered those things would be a different tool; this one is honest about being this one.
It reads from 10 to 100 and shows a word next to it, which makes it look like ninety-one settings, but it decides one thing: whether the object streams are used. At 40 and above they are, and that is where essentially all of the saving comes from; below 40 the file is written without them and the difference is usually negligible. The level is not a quality value and does not scale anything — it sits just above the point where it stops mattering, which is why the hint in the settings panel says so plainly.
A re-save can produce a LARGER file, and nothing here can prevent that: an image-heavy PDF that is already packed efficiently has nothing left to gain. What the page can do is refuse to hand you the worse file. When the produced bytes are not smaller, the row keeps the original, says so, and the download gives you the original back under its own name — the produced copy is measured and reported, but it is not what you receive. The size that matters on the row is always the size of the file you are about to get.
Nowhere, for the ordinary path. The compression runs in this tab, and closing it discards the result. The one exception is the optional import-by-URL feature, which fetches the document through this site's own server so that a browser can read a file from another host — a file brought in that way does travel, and any sentence claiming otherwise would be false for that path. Nothing in the code proves what the platform's analytics or error reporting can see at the page level, so the page claims what it can support and no more.
This one runs in your browser — the document is loaded and re-saved in this tab by pdf-lib, and the file never leaves your device on the ordinary path. Closing the tab discards the result. Two honest qualifications: the optional import-by-URL feature fetches the document through this site's own server, so a file you bring in that way does travel; and this page does not claim to be a zero-tracking or fully private environment, only that your PDF is not uploaded in order to be compressed.
No, and on this page that is the easy claim to make rather than the hard one. Nothing is re-encoded: the document is written out again with its content in object streams, and the images and fonts inside are copied exactly as they arrived. There is no quality level to lower, because there is no encoder to lower it. What does vary is how much smaller the file gets, which is a different question with a much less comfortable answer.
It depends on what the PDF is made of, and this page will not quote you a figure. The saving comes from packing many small object definitions into streams, so a document with a lot of them — a generated report, an exported form, a text-heavy file — can improve substantially, while a document whose weight is in its images gains little. If the images were already stored efficiently, the re-saved file can come out slightly larger, which is why the row keeps the original in that case rather than passing the bigger file on to you.
No, and here it is not merely that the field is missing — the mechanism cannot target a size at all. A target-size mode would mean discarding information until the number came down, and this tool discards nothing: it rewrites the container the document is stored in. The only control that exists is whether the object streams are used, and it moves the size by whatever the document happens to allow. Getting to a specific number would need the images re-encoded, which is a different tool than this one.
None of those. No pruning of leftover objects happens, no image is decoded or resampled, and nothing in the page touches colour. Fonts are copied rather than rebuilt or subsetted, the metadata is carried through unchanged, and there is no way to select which pages are kept — every page is kept, and the page count is displayed because the tool reads it, not because it offers a choice. Those are the features of a more invasive optimiser, and this page does not have them.
Not on the ordinary path — the document is compressed in this tab and is never sent in order to be compressed. The qualification is the optional import-by-URL feature: to read a document from another website, the fetch is made through this site's own server, so a file you bring in that way does travel through it before the compression happens in your browser. If you opened the PDF from your own disk, nothing was uploaded. The page does not claim to be a zero-tracking environment, only that your file is not sent away to be shrunk.
Because the re-save could not beat the original, and the page will not switch you over to a worse file to make the run look successful. When that happens the row keeps the original, shows its size, and says so — and the download gives you the original under its original name, because naming an untouched file '_compressed' would be a second untrue claim stacked on the first. It is common with scanned or image-heavy documents, and no setting on this page can change it.
200 MB per PDF, and PDFs only. Several can be queued and they are processed one at a time in the tab, each with its own row and its own result. An oversized file, or a file that is not a document at all, is declined with a message naming which of the two it was; if you are signed out, the daily conversion limit that applies across the platform applies here too.
