The compressed PDF got bigger
A PDF gets bigger after compression when the original was already vector text compressed with Flate. Our compression draws each page and re-encodes it as JPEG: on a text page, that image takes more room than the drawing instructions it replaced. In those cases, splitting or deleting pages is the right tool.
Verified 2026-08-12
- Dominant cause: the original was already compressed with Flate, the same algorithm as ZIP.
- Our compression rasterises: each page becomes a JPEG image, and the text stops being searchable.
- An A4 page at the light level is drawn with 2,005,644 pixels; at medium with 1,127,859; at strong with 607,185.
- A disadvantage of ours, said plainly: tools that recompress the images without touching the text layer do better here.
- With no loss at all: delete pages and split the file. Those are the first two things to try.
- The app detects the case and shows a different message, but the compressed file is still downloadable.
- This tool is part of the Premium subscription, EUR 1.99 a month, which removes the limits on every tool and adds OCR. Sign PDF opens on the free plan, with one signature a day. The EUR 1.99 a month subscription removes the limits everywhere and adds OCR.
Why did my PDF get bigger after compressing?
Because compressing is not squeezing the same content, it is swapping it for another representation. And a swap can come out worse.
A text PDF stores each page as drawing instructions: which font, which glyph, at which coordinate. That instruction stream is stored with the Flate filter, the same family as the deflate used in ZIP files. It has already been through a compressor, and it is remarkably compact: a thirty page text document weighing 400 KB spends about 13 KB per page.
Our compression replaces that with a photograph of the page. Each page is drawn onto a canvas and re-encoded as JPEG. A photograph of a white sheet with black letters cannot possibly be as cheap as the list of letters that produced it, because it is storing pixels instead of meaning. When the original was already compact, the swap makes the file fatter.
With a scan the opposite happens, and that is where the tool wins: the original was already a large image, and replacing it with a smaller image genuinely reduces the file, typically between 50% and 90%.
What are the concrete causes of a PDF growing?
Four of them, and they usually come in pairs.
| Cause | What happens | How to recognise it |
|---|---|---|
| The original was already vector text | Each page goes from Flate instructions to a photograph of the page | You can select words with the mouse in the original |
| The images were already compressed | JPEG re-encoded over JPEG adds artefacts and sometimes bytes | The original is an already optimised scan |
| The document has few pages | The fixed cost of one image per page never gets amortised | Fewer than five pages and under 1 MB |
| You picked the light level on a small file | Scale 2.0 draws 2,005,644 pixels per A4 page | You picked "Light, more quality" |
The second one deserves a note, because it is the least obvious. If your PDF is already a scan stored as good quality JPEG, decoding that JPEG onto a canvas and re-encoding it is a second generation: it loses quality and may not even save space, because the encoder is now coding the artefacts of the first compression as if they were detail in the document.
The fourth is the easiest to fix. The light level draws the page at twice the PDF base scale, which gives 1191 by 1684 pixels on an A4 sheet, that is 2,005,644 pixels per page. The medium level drops to 893 by 1263, which is 1,127,859 pixels, and the strong level to 655 by 927, which is 607,185. If you picked the light level on a document that was already small, try medium before concluding the tool is useless.
How do I know in ten seconds whether compressing will help at all?
Open the PDF and try to select a word with the mouse.
- If you can select individual letters, it is vector text. It will not shrink, and it may grow. Skip compression.
- If the cursor selects the whole page as one block, or selects nothing, it is an image. It will shrink a lot.
- If some pages let you select text and others do not, the document is mixed. Worth trying the light level and comparing sizes.
There is a second indicator, faster still: the ratio of size to page count. Under 100 KB per page is almost always vector text already at its minimum. Over 500 KB per page is almost always images, and there compression pays.
BestPDFInTheWorld compression rasterises. What does that cost you?
It costs you the document as a document, and that is a disadvantage of ours not worth hiding. After compressing:
- Text stops being selectable and searchable. Neither your PDF reader nor your operating system will find words inside the file again.
- Links and form fields disappear. What survives is the appearance, not the behaviour.
- Transparency becomes white, because the canvas is filled with white before the page is drawn.
- It is irreversible. Going back from image to text requires optical character recognition. We do have that tool (OCR), on the paid plan, but what it returns is a reading of the image, not the original text.
There is another class of compression that does none of this: instead of rasterising the whole page, it finds the images inside the PDF, lowers their resolution and recompresses them, leaving the text layer untouched. That is what Ghostscript and Adobe Acrobat do. In those cases text stays searchable and the file never grows on account of the text.
Compression at BestPDFInTheWorld rasterises: it turns every page into an optimised image and the text stops being selectable. There is no text preserving mode, and no hidden button to turn one on. If you need to keep the text searchable, a Ghostscript based tool serves you better; if the text does not need to be searchable, our compression gets it done on the spot, in your browser and without sending the file anywhere.
What do I do when compression is the wrong tool?
Ordered by what costs the document least.
| What to do | What you lose | When it fits |
|---|---|---|
| Delete pages | Nothing. What stays comes out intact. | There are pages you can drop |
| Split into parts | Nothing. Each part comes out intact. | The limit is per file, not per total |
| Recompress images only | Image resolution. The text survives. | You need searchable text |
| Compress (ours) | All text, links and form fields | The document is made of images |
Deleting pages and splitting cost zero quality and solve most real cases, because the problem is almost never that the document is heavy, it is that it is heavy against a specific limit. Sending six pages instead of eighty usually solves it on its own. Two 15 MB parts pass where one 30 MB file does not.
If the limit is an email one, count on less than the advertised number: the attachment travels base64 encoded and grows by 33%, so the 25 MB on Gmail fits a file of about 18 MB.
Does the app warn me when the file did not shrink?
It does, with a different message. When the result comes out the same size or larger than the original, instead of a reduction percentage you get: "Done, but this PDF was already light: it stayed at [size]. Compression pays off more on image PDFs."
Notice what that message does and does not do: it shows you the final size, but it does not say in words that the file grew. Compare the number on screen with the size of the original before saving the result. And keep the original either way: if the compressed file is no good, you still have somewhere to start again from.
The compressed PDF got bigger
Why did my PDF get bigger after compressing?
Because the original was already Flate compressed and our compression replaced it with a JPEG image of every page. On a text page, the image takes more room than the drawing instructions it replaced. It happens mostly on text documents with few pages.
Does compressing twice reduce it more?
No. The second pass redraws an already degraded image and recompresses it, which loses more quality and rarely saves space. If one pass was not enough, the route is deleting pages or splitting the file.
Which level should I use so the file does not grow?
Medium, which draws 1,127,859 pixels per A4 page, or strong, which draws 607,185. The light level draws 2,005,644 and is the one most likely to grow a document that was already small.
Is there compression that does not wreck the text?
There is, and it is not ours. Ghostscript based tools and Adobe Acrobat lower the resolution of the images inside the PDF without touching the text layer, so text stays searchable. Our compression rasterises the whole page and has no alternative mode.
The file grew. Have I used today's free compression?
Yes. The count is per operation run, not per useful result. Before compressing, do the mouse selection test: if you can select a word, the document is text and compression will not help.
What is the smallest a text PDF can get?
What it already is. A text PDF exported from Word or generated by a billing system comes out practically at its minimum, in the tens of kilobytes per page. If your document is in that range, there is no compression work to do.
Keep reading
How big a PDF a browser can take
The hard limits with numbers: byte block, canvas area, and the 384 MB ceiling in Safari on iOS.
PDF passwords: what actually protects
The two PDF passwords, the four ciphers, and why one of them comes off with a click.
I forgot my PDF password
The honest answer: without the open password there is no recovery. What actually works, and what is not worth trying.
Shrink it without wrecking the text
Deleting pages and splitting cost no quality at all, and they run in your browser.
Open the tools