Compress PDF Files Online
Bring a heavy PDF down to a size that clears an attachment cap or a filing portal, without losing the detail you still need to read.
Upload or drop document to translate
Max. file size 1 GB
Almost all of the size is in a few objects
Weight in a PDF is never spread evenly across the pages. Text is tiny. A page of prose costs a few kilobytes because it is stored as a list of characters and positions. What makes a file heavy is almost always one of a short list of things, and knowing which one you have decides what compression can do for you.
- Photographs placed at full camera resolution and then scaled down on the page. The whole original is still in the file, even though only a fraction of its detail is visible at that size.
- Scanned pages, where every sheet is one large image. A scan is the heaviest kind of PDF by a wide margin, and its size scales with the resolution the scanner was set to.
- Whole fonts embedded rather than subsets. A font carrying every glyph in its family costs far more than one carrying the few hundred characters the document actually uses.
- Leftovers from incremental saves. Editing and saving a PDF repeatedly can append each revision instead of rewriting the file, so old versions of pages you already deleted may still be inside it.
- The same logo or letterhead image stored once per page instead of once per document.
A quick way to guess before you upload: divide the file size by the page count. A text document sits in the tens of kilobytes per page. Anything in the megabytes per page is images, and that is where the reduction will come from.
Two very different operations behind one button
Compression tools bundle two kinds of work together. One is free of consequence and the other is permanent, and the reason results vary so much between files is that different documents have different amounts of each available to give.
Lossless: nothing is thrown away
Streams inside the file are repacked with a more efficient encoder, embedded fonts are cut down to the characters actually used, objects that nothing references any more are deleted, identical images stored several times are collapsed into one, and superseded revisions from earlier saves are dropped.
Every pixel and every letterform in the output is identical to the input. On a text heavy document that has been edited many times this alone can be a large saving. On a clean, freshly exported file it may save almost nothing, because there is no waste to remove.
Lossy: detail is given up for size
Images are reduced to a lower resolution and re-encoded at a chosen quality setting. Information that is discarded at this stage is gone from that copy for good, and compressing the result a second time makes it worse again rather than smaller in any useful way.
This is where the dramatic reductions come from, and it is also the only stage that can damage a document. A file that is mostly text has hardly any lossy saving to offer. A file of scanned pages is almost entirely image, so it responds strongly, which is exactly why scans need the most care.
What heavy compression costs a scan you still have to read
Character recognition works by matching letterforms, so it depends on clean edges. Lossy image compression attacks precisely those edges: it works in blocks and leaves faint ripples around sharp transitions between dark and light. On a photograph that is invisible. On a page of small print, where every letter is a sharp transition between dark and light, it is the difference between a word being read correctly and being read as something else.
Resolution matters just as much. Roughly 300 dots per inch is the usual recommended minimum for reliable recognition of body text, and small print, footnotes and stamped detail want more rather than less. Downsampling a 300 dpi scan to fit a size limit is the single most common way a document becomes unreadable by software while still looking acceptable to a person on screen.
The order of operations solves it. Recognise the text first, while the pages are still at full quality, so the words end up stored as text in the file. Compress afterwards. The images can then be squeezed hard, because the searchable and translatable content no longer depends on them. DocTranslator reads text inside scanned documents and image files, so uploading the good copy rather than the compressed one gives a noticeably better result.
When a PDF will not get much smaller
It has already been compressed
Files exported from modern software usually arrive with their images already encoded efficiently and their fonts already subset. There is no slack left to take up, so a second pass returns a file of roughly the same size, or a slightly worse looking one at the same size.
It is mostly text
A long contract or a manual with no photographs is small to begin with. If a text document is unexpectedly heavy the cause is usually embedded fonts or accumulated save history rather than page content, and the lossless pass is the part that helps.
It is vector artwork
Engineering drawings, maps and detailed charts can hold hundreds of thousands of individual lines and shapes. That is data, not pixels, so there is nothing to downsample. Reducing it means simplifying the drawing, which changes what it shows.
It is encrypted or signed
A password protected file has to be unlocked before anything can rewrite its contents. A digitally signed file can be compressed, but the rewrite changes the bytes the signature covers, so the signature no longer validates. Compress before signing rather than after.
The limit you are trying to clear decides how hard to push
Nobody compresses a file for its own sake. There is always a number on the other side: an attachment cap on a mail server, an upload field on a government or court portal, a supplier system that rejects anything above a set size, a storage quota. Find that number before you start, because it tells you whether a lossless pass is enough or whether image quality has to be spent.
When the target is far below what a single compression pass can reach, stop pushing quality down and change the shape of the job instead. Cutting the document into parts with Split PDF and sending them separately keeps every page readable, where forcing one file under an aggressive limit does not.
Upload size is rarely the constraint on our side. DocTranslator accepts files up to 1 GB or 5,000 pages on Monthly and Annual plans, so a document usually only needs compressing for wherever it goes next, not for translation.
Translate from the good copy, send the small one
Where a document is both going to be translated and going to be sent somewhere with a size limit, do the translation from the original. Quality of the source is what the recognition step has to work with, and a file that has already had its images cut down gives it less to read.
The translated output tends to be lighter than the scan it came from anyway, because words that were pixels in the original are text in the result. A scanned page holds an image of every letter, while a translated page holds the letters themselves, and that alone changes the arithmetic.
Keep the untouched original somewhere regardless. A compressed copy cannot be restored to its earlier quality, and the day someone asks for a clean version of a certificate or a signed page is the day you find out whether you kept one.
Check the smallest print
Judge a compressed file on the hardest thing in it, not on the first page. Zoom in on a footnote, a stamp, a handwritten signature or a table of figures. If those are still legible the compression was safe; if the numbers have gone soft, take a step back on the quality setting. Reviewing at the zoom level a reader will actually use, rather than a page fitted to the screen, is what tells you the difference.
How to compress a PDF properly
Work out what the target size is
Find the limit you have to clear first, whether that is a mail attachment cap or an upload field on a portal. It decides whether a safe lossless pass is enough or whether image quality has to be spent.
Recognise the text before you shrink anything
If the pages are scans, get the words out of the pixels while the images are still at full quality. Once the text is stored as text, the images can be compressed hard without costing you anything you need.
Upload and let the file be rewritten
Create a free account and upload the document. Unused objects, duplicated images and oversized font sets come out first, which on an edited file is often enough on its own.
Inspect the hardest page, then keep the original
Zoom in on a footnote, a stamp or a column of figures to confirm the result is still legible, and archive the uncompressed source. Detail given up in a lossy pass cannot be recovered later.
Compressing a PDF: common questions
Why did my file barely get smaller?
Because there was little to remove. Text costs very little to store, so a document without photographs is already close to its floor, and a file exported by modern software usually has its images encoded efficiently and its fonts already cut down. Big reductions come from image data, which means scans and photographs.
Will compressing damage my document?
The lossless part cannot, since it only removes waste and leaves every pixel identical. The lossy part reduces image resolution and quality permanently, and how much that matters depends on what the images are. A photograph tolerates it easily, a page of small print does not.
Should I compress a scan before running OCR on it?
No, do it the other way round. Recognition matches letterforms, and lossy compression blurs exactly the sharp edges it relies on while downsampling removes the detail small print needs. Recognise the text at full quality first, then compress, because by then the words are stored as text and no longer depend on the images.
What resolution should a scanned page be?
About 300 dots per inch is the usual recommended minimum for reliably reading body text, and anything with small print, footnotes or stamped detail benefits from more. Dropping below that to hit a size limit is the most common way a file stays readable to a person on screen while becoming unreadable to software.
Can I compress the same file twice to make it smaller again?
It rarely helps and often hurts. The lossless savings were already taken on the first pass, so a second run mostly re-encodes images that have already lost detail, which degrades them further without a meaningful reduction. If one pass did not reach your target, split the document instead of squeezing it again.
Why is my PDF huge when it is only a few pages?
Divide the size by the page count. Megabytes per page means image data: photographs kept at full camera resolution but displayed small, or scanned pages captured at a high setting. Text pages sit in the tens of kilobytes, so a heavy text document usually points to fully embedded fonts or leftover revisions from repeated saves.
Do I need to compress before uploading for translation?
Usually not. DocTranslator accepts files up to 1 GB or 5,000 pages on Monthly and Annual plans, so size is rarely the constraint here, and a higher quality source gives better results when pages have to be read from images. Compress for wherever the document goes afterwards, not for the translation itself.
What happens to a signed or password protected PDF?
A protected file has to be unlocked before its contents can be rewritten at all. A signed file can be compressed, but rewriting it changes the bytes the signature was applied to, so the signature stops validating. Where a document needs both, compress it first and sign the smaller version.
Get your PDF under the limit
Strip the waste out of a heavy file, keep the small print legible, and translate the document into any of 120+ languages once it is the right size.
Our Partners
More PDF Tools & Translation Services
PDF Tools