In-Depth How-to Guide
#pdf compression#large files#scanned documents#browser limits#zip parts

Compress a 300 MB Scanned PDF Without Freezing the Tab

DocsSeva Team6 min read
A browser showing a 312 MB scanned PDF compressing page by page, with a ZIP of numbered part files in the Downloads bar.

Key takeaways

  • Structure modes parse the whole file at once, which is what freezes the tab on very large PDFs.
  • Above about 200 MB the job is routed to the page-by-page image engine automatically.
  • Most of the saving comes from capping the longest side, not from the JPEG quality setting.
  • More than 40 pages of rasterised output arrives as a ZIP of numbered part PDFs.
  • Split out the pages you actually need before compressing; it is faster and gives one clean file.

The land records came back from the scanning shop as one PDF: 180 pages, 312 MB, every page a colour photograph of a sheet of paper. The registrar's portal takes 25 MB. You open a compressor, drop the file in, and thirty seconds later the tab goes white with "Page Unresponsive".

That crash is not really about compression. It is about how the file gets read in the first place.

Why a very large PDF freezes the tab

The gentler compression modes work on the PDF's structure — they open the document, find the images inside it, re-encode those, and write the file back out with every character of text still selectable. To do that, the library has to parse the entire document before it can touch anything.

For a 4 MB file, that parse is instant. For 312 MB it is a single long block of work on the browser's main thread, which is the same thread that draws the page and responds to your clicks. The browser cannot tell the difference between "busy" and "hung", so after a few seconds it offers to kill the tab.

The PDF Compressor avoids this by not going down that road at all above a certain size. Past roughly 200 MB, whichever level you picked, the job is routed to the image engine instead — the one that works page by page and hands control back to the browser between batches of pages. You are told this happened; you did not misclick.

What the image engine actually does to your pages

It renders each page as a picture and rebuilds the document from those pictures. Two settings decide the outcome, and they differ by the level you chose:

LevelLongest side capped atJPEG quality
Low1900 px72%
Medium1600 px60%
High1050 px45%
Screen1600 px60%

Most of the saving comes from the first column, not the second. A shop scanning at 600 DPI produces an A4 page around 5000 px on the long side; bringing that to 1600 px throws away about 90% of the pixels before a single quality setting is applied. Your text stays perfectly readable on a screen at 1600 px. It stops being good enough to reprint at full size and then rescan, which is a trade worth making consciously.

For a 180-page scan, Medium or Screen is usually the right call. High is for when the portal limit is brutal and legibility on a phone screen is all you need.

Why your download arrived as a ZIP

This surprises people, so here is the reason.

Holding 180 rebuilt pages in one document and then serialising all of them in one go is exactly the operation that runs the tab out of memory. So the output is assembled in bounded chunks of 40 pages. Each chunk becomes its own small PDF, is written out, and is released before the next chunk starts.

When a job produces more than one chunk, you get them together as a single ZIP named compressed_ plus your file's name, containing yourfile_part-01.pdf, yourfile_part-02.pdf and so on. Our 180-page job comes out as five parts.

Before you rush to stitch them back together, check what the portal actually wants. Plenty of government upload forms have a per-file cap and are perfectly happy with "Document 1 of 5" — and five 12 MB files clear a 25 MB limit while one 60 MB file does not. If you genuinely need a single document, PDF Merger will join the parts, but you are re-creating the large single file the split existed to avoid, so expect it to be heavy.

The trade you are making, stated plainly

Rasterised pages have no selectable text. None. You cannot search the output, you cannot copy a survey number out of it, and neither can anyone you send it to.

For a scanned document this often costs nothing, because a photograph of a page never had selectable text to begin with. But if your PDF had been through OCR earlier and carried a text layer, that layer is gone after this. Two ways round it:

  • Run PDF OCR on the compressed file afterwards. It reads the page images and writes an invisible text layer back on top of them, so the document becomes searchable again without changing how it looks.
  • Or keep the compressed copy strictly for uploading, and keep your searchable original for your own records. Naming them clearly is cheaper than regretting it in eight months.

Numbers to expect, and when to stop

Screen mode publishes a range of roughly 55% to 85% smaller. On our 312 MB file, that is somewhere between 47 MB and 140 MB. Real scans of typed pages sit at the good end of that; scans with photographs, stamps and coloured letterheads sit at the poor end.

Which means the portal's 25 MB limit may still not be met, and no amount of re-running will fix it. Compressing an already-compressed file gives you very little the second time and costs quality — the tool detects its own previous output through the producer metadata it writes, and warns you rather than letting you grind a file down to mush.

At that point the answer is fewer pages, not more compression.

Do the cheap step first

This is the part I would tell a colleague before anything else on this page.

If the registrar needs pages 41 to 52, do not compress 180 pages. Pull the range out with PDF Splitter first, then compress twelve pages. It finishes in a fraction of the time, comes out as one clean PDF instead of a ZIP of parts, and the result is small enough that the compression level barely matters.

The same logic applies to the blank sheets a duplex feeder adds behind every single-sided page. Half of a 180-page scan is sometimes nothing but backs of pages.

Two more things worth knowing before you start

The preview disappears on big files. Above about 50 MB the on-page preview is skipped and you get a card telling you so. Compression still runs on the whole file; only the thumbnail is skipped, and rendering page one of a 312 MB document just to show it to you is not worth the wait.

You cannot get a bigger file back. If the rebuilt output — or the ZIP of parts — is not smaller than what you started with, the original is returned unchanged. That is what protects text-heavy and vector-heavy PDFs, which genuinely do come out larger when every page is turned into a photograph.

Batches of ordinary-sized files are a different job with a different shape: up to 20 PDFs at a time, all processed in your browser, returned together. The rest of the size-reduction tools sit under compressors.

DocsSeva Team

Published on August 24, 2026

Open the Compressors

Free, no sign-up, and your files are never stored.

Open the Compressors

In-Depth How-to Guide

Portal Says Invalid File Type but Your JPG Looks Fine

A renamed PNG is still a PNG, and upload forms check the bytes. How to confirm what your file really is, which mismatches are harmless, and how to fix a real one.

DocsSeva Team7 min