Compress PDF
What Compress PDF actually does
Most compressors give you a slider and a smaller file, and you find out afterwards whether your text survived. Rasterizing a text document is the wrong operation for it: measured here, a text-only PDF came back 16 to 62 times larger with zero selectable characters. This workbench offers two passes and says which one it is doing. The keep-text pass changes only the image XObjects, so text, vector artwork, links, form fields and the outline are untouched, and it reports every image it found, every image it replaced, and a named reason for every image it left alone. The rasterize pass is the older behavior, kept for scans, and it still states that the result is image-only before it will hand you anything.
How to use it
- Choose one PDF within the displayed byte and page limits; its text, pages, document features, and raster-image paint operators are inspected locally.
- Choose a pass. Keep text re-encodes the photographs in place and leaves the page content alone; rasterize turns each page into one JPEG and says so.
- Review the per-image account or the same-page comparison, then download the PDF and a receipt whose every byte figure was read back out of the file that was written.
Useful for
- Get a photo-heavy report under an email limit without losing the ability to search or quote it.
- Reduce a scanner-created packet where there is no text to preserve.
- Show someone exactly which images changed, and by how much, rather than asserting that the file is smaller.
Limits worth knowing
- The keep-text pass re-encodes JPEG images only. A PNG-style FlateDecode image, a stencil mask, an explicitly masked image, an image already within the resolution budget, and an image whose re-encode is not smaller are each left alone and reported by name.
- Each image is capped at the page width in inches times the target resolution, which is the widest it could usefully be on that page rather than its true placed size, so an image drawn small keeps more pixels than it needs.
- The rasterize pass produces an image-only derivative: selectable text, vector scaling, links, forms, bookmarks, tags, attachments, layers, signatures, and encryption are not preserved.
- The browser accepts one PDF up to 50 MiB and 60 pages, with an 8,192-pixel canvas-side cap, 16 million pixels per page, 80 million pixels per run, and a 100 MiB candidate cap.
- A candidate that is not smaller is reported as no size win and cannot be downloaded as a compressed result.
Questions people ask
Will my text still be selectable?
With the keep-text pass, yes. It replaces image data only and never touches the page content stream, so the character count is the same before and after and the workbench shows you both numbers. The rasterize pass deliberately does the opposite and says so before it builds anything.
Why did it not change some of my images?
Because it will only re-encode what it can re-encode honestly. Every image it declines is listed with a reason: an unsupported filter, a stencil or explicit mask, an image already small enough for the target resolution, an image below the minimum size, or a re-encode that did not come back smaller.
What happens if the new PDF is larger?
The workbench shows the byte figures and the verification checks but does not enable the PDF download or call the run successful. Try the other pass or another setting, or keep the source.
Is the PDF uploaded?
No. After the route loads, diagnosis, re-encoding, PDF creation, comparison, hashing, verification, and downloads happen inside this browser tab.