ALL PDF TOOLS37
Recompress PDF Images
Recompress PDF Images — Shrink a Photo-Heavy PDF
Re-encode the images inside a PDF at a lower quality to make the file smaller — lossy, irreversible, and entirely in your browser.
About Recompress PDF Images
Some PDFs are large because of how they are built, and some are large because of what is inside them. This tool is for the second kind. It walks every image the document embeds, decodes the ones stored as JPEG, redraws them and writes them back at a lower JPEG quality — and, when you give it a target size, at smaller pixel dimensions as well. That is a lossy edit and it cannot be undone, so keep your original. What it never touches is the text and the vector artwork: those come out exactly as they went in, which is why a report with photographs in it stays sharp everywhere the photographs are not. There are two ways to drive it. Quality offers three presets and makes a single pass at the one you pick. Target size takes a number in KB and searches for it — quality first, then progressively smaller image dimensions, rebuilding the document each time and keeping the best-looking result that still fits under your number — the highest quality that lands under it, not the smallest file it can make. When nothing fits, you get the closest size it managed and a plain statement that the target was not reached, rather than a file quietly larger than you asked for. The more interesting half is what it refuses to do. Only images stored as JPEG are eligible. An image held as JPEG 2000, CCITT fax, JBIG2 or a deflated bitmap is written back byte for byte as it came in, and so is anything carrying a transparency mask, anything with its own decode array or predictor parameters, anything in CMYK or an indexed or ICC-tagged colour space, and anything that is not eight bits per channel. Rewriting those needs codecs and colour handling this tool deliberately does not carry, and a wrong guess prints an inverted or falsely coloured page. Every skipped image is counted, grouped by reason and shown on screen, because a compressor that silently does nothing is worse than one that says so. There is a per-image guard too: if the re-encode comes back bigger than the image it replaced, the original image stays. Each image is processed once however many pages draw it, and the finished document is written back with its objects packed into object streams. If you would rather not lose anything, try Compress PDF first — it repacks the same file without touching a single pixel.
Questions
How is this different from Compress PDF?
Compress PDF is lossless. It drops objects nothing references, deflates streams that arrived uncompressed and repacks the file, and every pixel and character comes out identical — it has no image encoder at all, which is why a PDF that is mostly photographs often comes back the same size to the byte. This tool is the lossy counterpart: it decodes the JPEG images the document carries and writes them back at a lower quality, which is the only way a photo-heavy file gets meaningfully smaller. Try Compress PDF first if losing nothing matters, and this one when the size matters more than the pixels.
Which images does it skip, and why?
Only images stored as JPEG are re-encoded. Anything else is written back byte for byte as it came in: JPEG 2000, CCITT fax, JBIG2 and deflated bitmaps, images carrying or acting as a transparency mask, images with their own decode array or predictor parameters, CMYK, indexed and ICC-tagged colour spaces, and anything that is not eight bits per channel. Each of those would need a codec or a colour conversion this tool does not carry, and guessing at one prints an inverted or falsely coloured page. Every skipped image is counted and its reason named on screen, so you always know how much of the file was actually in play.
Can I get the original image quality back afterwards?
No. The re-encode is destructive and nothing in the output remembers the pixels that went in — detail a lower-quality JPEG pass throws away is gone, and running the result through again only compounds it. Keep your original file. Since the work happens on your own machine, the copy on your disk is the only copy that exists.
Why didn't it reach the target size I typed?
Because some documents cannot get there. The search lowers JPEG quality first and then image dimensions, rebuilding the whole file each round, but text, fonts and vector artwork are never re-encoded and neither are the images it had to skip — so there is a floor below which the document will not go. When the run ends above your number it says so and reports the closest size it reached, rather than handing you a download that quietly missed the target. If every image was skipped, the only saving left is the structural one — the file is still repacked into object streams, which usually moves it a little — and it tells you that is all that happened.
Will the text and line art get fuzzy too?
No. Only the embedded images are replaced. The text, the fonts it is set in and the vector drawing are never re-encoded, so a page of type comes out identical after a run — the loss is visible only where there was a photograph to lose it in.
Does it help with a scanned PDF?
Usually, and that is the case it exists for: most scanners store each page as a JPEG, which is precisely what this tool re-encodes, so a multi-megabyte scan can lose a large share of its size. Two kinds of scan will not budge — a bilevel fax-style scan stored as CCITT or JBIG2, and one whose pages sit in a colour space this tool skips. Both are reported as skipped rather than quietly passed through.
Is my file uploaded to a server?
No. Transmute processes everything locally in your browser using JavaScript and WebAssembly. Your files never leave your device — there is no server, no upload, no cloud processing.