How to compress a PDF without losing readability

To compress a PDF without losing readability, upload it to “Compress PDF” and pick an option: High quality, Balanced or Smallest size, or switch on Fit a size limit and name the limit. Then compare pages before and after, and check that text and stamps are still legible.

How to compress a PDF without losing legibility

The intake form says forty megabytes and your file is fifty-two. Compression with Compress PDF solves that in a minute: you choose how hard to squeeze the file, or simply name the size it has to fit into.

It helps to understand up front where the weight actually goes. Text in a PDF weighs almost nothing. What makes a document heavy is images: scans, photos, screenshots, logos left at their original resolution. So compression works mostly on images, and on a contract made of plain text it will not do much.

Three options instead of levels and profiles

Above the run button sits a "How to compress" block with three options. The middle one, Balanced, is selected by default.

OptionWhat happens to the imagesWhen to use it
High qualityResampled to about 200 pixels per inch, light compressionPrinting, documents with small stamps and signatures
BalancedAbout 150 pixels per inch, noticeable compressionEmail, upload portals, reading on screen
Smallest sizeAbout 110 pixels per inch, strong compressionTight limits, sending through a messenger

Text and vector graphics are not degraded by any of the options: only the pictures inside the pages are resampled. That is why on a contract with no scans the options barely differ, while on a stack of phone photos of a document the difference is several times over.

If in doubt, start with Balanced. It keeps small type and stamp impressions legible, and the weight usually drops several times. You can always adjust the result afterwards without uploading the file again.

Fit a size limit: when a specific number matters

Often you need not "smaller" but a specific figure: a government portal accepts up to 5 MB, a mail server bounces attachments over 20 MB. That is what the Fit a size limit switch is for. Turn it on and pick the limit: 2, 5, 10 or 20 MB, or type your own under Custom.

The service then tries the compression steps itself and returns the best quality that still fits the limit. If the file is already under the limit, it comes back unchanged: there is no reason to degrade something that already passes. The limit applies to each file separately, so in a batch every document gets its own degree of compression.

Sometimes the limit is out of reach. Below a certain point the text in a scan stops being readable, and the service does not go there: instead of an unreadable file you get a message saying it cannot get below N MB without losing readability, with suggestions for what to do next. Converting the scan to black and white, removing unneeded pages or splitting the document usually helps.

What sits under Advanced

Below the options there is a collapsed Advanced section. Most people never need it, but it helps in three cases.

Image resolution. The default is Auto: the resolution follows the selected option. You can set it explicitly to 300, 200, 150 or 100 pixels per inch. 100 is the minimum at which ordinary text in a scan still reads; 300 keeps pictures fit for print, but the file will be heavier.

Don't re-encode images. This checkbox leaves the pictures as they are and compresses only the structure of the file: fonts, internal data, duplicate objects. The gain is smaller, but not a single pixel changes. It is the right choice for a document headed to a court or an expert examination, where image fidelity matters. The compression option makes no difference then.

File name. Set the name of the result here, so you do not have to rename it after downloading.

Step by step

1. Open Compress PDF and upload the document. A line with its size, page count and kind, such as "scan", appears under it. 2. Pick an option. Balanced suits most tasks. 3. If there is a hard limit, switch on Fit a size limit and enter it. 4. Start the compression and wait for the result: it shows the size before and after. 5. Open Compare the quality and use the zoom on the busiest page rather than the first: the one with small type, a stamp or a table.

Step five matters more than it looks. A title page almost always survives compression well; what falls apart is usually the small caption under a table on page eight.

Compare and adjust without starting over

Under the result there is a Compare the quality block: the same page before and after compression side by side, with a zoom that magnifies the same spot on both. You can see whether a stamp or a small digit in the bank details has blurred before you even download.

If the result is not right, there are two buttons next to it. Compress harder takes the next step towards a smaller file, Restore quality takes a step back. No need to upload again: the service reuses the source it already has. The first three such retries of one document are free; after that each counts as a regular operation, and the text under the buttons says so. With Fit a size limit only Compress harder is offered: it lowers the limit by about a quarter, since the quality is already the best the chosen size allows.

Two more things the service tells you before you start. If the document has form fields, they will be kept. If it carries a digital signature, the signature will no longer be valid after compression, and the compressed file has to be signed again.

When a file does not shrink

This happens, and it is not a failure.

If no pass produced a file smaller than the source, the service returns the original unchanged and tells you so, saying that the file is already compressed and cannot be made smaller without losing quality. The logic is simple: handing back a file that got heavier or stayed the same but lost quality would be worse than saying honestly that there is nothing to compress here.

There are usually two reasons. The document is already compressed: it was exported from a system that optimises on output, or it went through compression before. Or the document simply has no images, and text and vector graphics shrink poorly.

If the weight still has to come down, what helps next is not compression but less content: remove surplus pages with Remove pages, or split the document into parts with Split PDF and send them separately.

Why the result is never heavier than the source

This is engineered deliberately, and knowing it is useful.

Compression does not make a single pass. It tries several processing variants and at the end compares them against each other and against the original file. With a compression option chosen, the smallest of them is what you get; with Fit a size limit, the best-quality one that fits the limit. Either way the result is no heavier than the original: if no variant came out lighter, the original wins.

A practical conclusion follows: trying compression risks nothing. The worst that happens is your own file coming back with an honest message. The "I compressed it and it got bigger" situation cannot occur here by construction.

The same fact means there is little point cycling through settings hoping for a miracle. If Smallest size produced nothing, High quality certainly will not: it recompresses the images more gently, and without images there is little left to compress. What needs changing is not the settings but the approach, meaning less content.

Several files in one go

You can compress a batch of documents at once, and the chosen option applies to all of them equally. With Fit a size limit the limit applies to each file on its own, so ten scans of uneven quality each get their own degree of compression.

The output is an archive whose files carry your original names with a `compressed_` prefix. If two uploaded files shared a name, numbers are appended so nothing gets overwritten.

The report on the result is not a lump sum, though: the service names the individual files it could not shrink. That matters for a batch, because the total archive weight can fall thanks to the successful files and hide the one stubborn document that still misses the limit.

If the set is mixed, split it into groups. Text documents barely shrink under any option, while on scans the choice of option is where the entire gain comes from.

One more way to shed weight: drop the colour

If compression squeezed out all it could and the file is still heavy, there is a trick people forget.

A colour scan of a black-and-white document stores colour that is not really there: the greyish tint of the paper, the yellow of the lamp, the bluish shadow of a fold. All of it takes up space alongside the meaningful data. Converting to greyscale with PDF to Grayscale removes that ballast entirely, and on a batch of scans the gain is usually larger than from compression itself.

There are three greyscale profiles, and the default is the standard one, safe for mixed office documents. Caution is due where colour carries meaning: a blue stamp, a coloured signature, a line picked out in highlighter. After conversion they stay legible but stop being coloured, and if the receiving side expects colour, this route does not suit.

For the best quality, greyscale first and compress second: compression then spends no effort on colour data you were going to throw away anyway. If the file is already compressed and still too big, converting the compressed result to greyscale helps too: that is the step the service suggests under the result.

What is pointless to compress

A document made of text and tables shrinks by a few per cent. Expecting a meaningful win there is unrealistic, and if such a file misses the weight limit, the cause is probably the page count rather than the pictures.

A separate case is a file that has already been compressed. A second pass barely reduces the weight while quality drops a second time, because the artefacts of the previous pass compound with the new ones. If the result disappointed, go back to the source and compress it once with different settings rather than running the finished file through again.

A scan made at low quality to begin with has little left to compress: there is not much data in it. Such a document is easier to reshoot than to optimise.

A test on five practice PDFs: sources and results

On 21 October 2026 we ran the current compression handler on five synthetic documents. On 22 October we repeated all fifteen runs after fixing how page orientation is preserved. This is a local technical experiment, not a measurement of the site's speed, statistics on users' files, or a comparison with other services. The set contains no real passports or client documents. The colour scene was drawn programmatically; the scans contain test lines, digits and a mock stamp.

Practice fileBefore, bytesAfter Balanced, bytesReduction
Text PDF with unoptimised streams83,2694,51394.58%
Synthetic colour scan9,234,714963,82289.56%
Greyscale scan4,080,871407,22790.02%
Synthetic colour scene3,254,283389,57888.03%
Small optimised PDF1,12594316.18%

The numbers show why the bare promise "we will shrink any PDF several times over" is not enough. The text sample was deliberately saved without stream optimisation, so its gain is unusually large. In the small optimised sample the gain came to 182 bytes. The synthetic noise in the scans does not reflect the variety of real documents either; these percentages cannot be carried over to your file.

The source size of the colour practice scan is 9,234,714 bytes. With a 2 MB target the result took 1,521,404 bytes, and with a 5 MB target 3,251,187 bytes. Both results fit under the ceiling that was set, but the 5 MB mode kept more image data. The Balanced option produced yet another size: hitting a target limit and shrinking as far as possible are different tasks. If a document already fits the target, the handler may limit itself to light optimisation.

A megabyte in this experiment is 1,048,576 bytes, which is how the handler counts. If the recipient's form uses decimal megabytes, compare bytes or leave some headroom. Neither fitting under a limit nor successful processing by itself proves that the document will be accepted.

We reopened every resulting PDF and rendered all of its pages, then checked the page count, the geometry and whether the extractable text was preserved. For the scans, this text check does not replace reading the image: they have no text layer to begin with. First-page comparisons help show changes but do not guarantee that any given fragment stays legible in print. Check small details at the scale you need.

You can download the results table as CSV, the full protocol with parameters and SHA-256 and the methodology for repeating the experiment. The methodology lists the source PDFs and the result of each mode; the practice files may be reused under CC0. To check for yourself, you can take the colour practice scan. For a specific limit, use the pages for compressing to 2 MB or to 5 MB, and compare the result with the source after processing.

Author:

How much can I compress a PDF before quality drops?

Scanned documents usually shrink 2-5x with no visible loss, because the images inside the pages carry most of the weight. A text PDF exported from Word is already lean, so there is little left to compress.

Will compression ruin stamps and signatures on scans?

Not at moderate settings. Aggressive compression can blur small stamps and handwritten signatures, so after compressing, open the file and check those exact spots.

Why is my PDF still large after compressing?

It usually holds full-resolution photos or a 600 dpi scan. The Smallest size option or a lower limit under Fit a size limit helps, and for long documents, splitting the file or removing unneeded pages helps more.

Do I need an account?

No. The basic workflow is available without creating an account.

How long are files stored?

Files are used only for the selected operation, kept for a limited time and deleted automatically. We do not use uploaded documents to train AI models.

How many PDF pages does 150 MB hold?

Depends what's on the pages. A text-only PDF at 150 MB could run into the thousands of pages, since plain text barely adds weight. A high-resolution scan, though, can weigh several MB per page, so 150 MB may hold only a few dozen scanned pages. The real number depends on scan resolution and whether it's a photo or plain black-and-white text, and it's more reliable to compress the file once and check the result than to estimate ahead of time.