PDFGym
EN

See what makes your PDF large

or drop one here
Choose a PDF first.

A PDF that weighs 48 MB can hide one giant photo, a dozen full font embeds, or ten forgotten revisions. This page opens the file and sorts every byte into eight buckets, from images down to leftover file wrapping. The totals reconcile to the exact size on disk, the report changes nothing, and the file never leaves your device.

The 48 MB portfolio that would not email

When a file refuses to fit an upload cap, guessing at the cause wastes the afternoon. Measure first, then fix only the part that holds the bytes.

Take an event photographer sending portfolio-2025.pdf to a gallery: 36 pages, 48 MB, and the submission form stops at 25 MB. The breakdown shows Images at 44 MB with a median placed resolution near 420 DPI, plus one 9 MB panorama sitting on page 7. The Next step panel names oversized images as the clearest opportunity and offers a single button.

One press on Compress this PDF carries the file straight into Compress PDF without uploading it, and running the breakdown again afterwards shows whether the fix moved the right bucket.

Four passes from file to finished report

  1. Open the PDF

    Press Choose a PDF, or drop the file on the page. Its name and size appear under the picker to confirm the right file loaded.

  2. Run the analysis

    Press Analyze PDF size. The Reading your PDF progress walks the document graph, matches objects to the pages that use them, and reconciles every byte.

  3. Read the chart and the table

    Where the bytes go stacks the eight buckets with object counts and percents, above a Largest objects table of the top 10 with Object, Type, Location, Details, and Stored size columns. Image rows carry dimensions, placed DPI, and codec, and locations read Page 7, 3 pages, or Document.

  4. Follow the single recommendation

    The Next step panel names one move with its button. With images dominant it offers Compress this PDF, which opens compression with the same file already loaded and nothing uploaded. The proof line reports every byte accounted for with zero unexplained.

Why the totals always match the file size

Each figure is the stored byte count, never the decoded one. A JPEG can unfold into many megabytes of pixels while occupying a fraction of that on disk, so the report states what the file carries. Parsing normalizes object syntax along the way, which is why measured sizes get apportioned back to the physical file instead of being quoted raw.

Headers, cross-reference data, and trailers keep their own File wrapper bucket rather than dissolving into the rest, and the eight buckets must add up to the source length exactly. When they refuse to reconcile, the run errors out and points at Repair PDF instead of printing a wrong total. Buckets with zero bytes stay out of the bar, and the Largest part badge ignores the wrapper so the headline always names real content.

The eight buckets every byte lands in

Run the report first, then read the row that matches your tallest bar. The right column says whether squeezing the file can shrink that bucket.

The eight size buckets and whether compression can shrink them
BucketWhat lands in itCan compression shrink it
ImagesWhat lands in itPhotos, scans, and logos at stored size, never decoded pixel size. Can compression shrink it Yes Compress PDF downsamples placed pictures to 150 DPI and re-encodes them.
FontsWhat lands in itEmbedded font programs. A full embed costs the same for 12 glyphs or 12,000. Can compression shrink it TrueType only Compression subsets embedded TrueType to the glyphs on the pages. Type 1 and CFF programs stay untouched.
Page contentWhat lands in itPer-page drawing instructions: text runs, vector drawings, placements. Can compression shrink it Rarely Big values here mean vector drawings, since text instructions stay tiny, and these streams already arrive compressed.
Embedded filesWhat lands in itFiles attached within, like source data sitting behind a chart. Can compression shrink it No Only deleting the attachment removes these bytes.
MetadataWhat lands in itDocument info dictionary, XMP packet, and editing history. Can compression shrink it No Remove PDF metadata wipes it for privacy. For size it is too small to chase.
Unreferenced objectsWhat lands in itLeftovers from earlier saves that the current pages no longer reach. Can compression shrink it Often A fresh save may clear them, and may also void a digital signature.
PDF structureWhat lands in itOutlines, annotations, form definitions, and the page tree holding them. Can compression shrink it No The fixed cost of the document itself, independent of how it was exported.
File wrapperWhat lands in itHeaders, cross-reference tables, and trailers around the objects. Can compression shrink it No The binding around the pages. Files that grew through many small saves show more of it.

The report only measures. It never rewrites the PDF.

Readings that confuse first-time users

A Fonts bucket near a third of the file looks like easy savings until compression barely moves it. The subsetter only trims embedded TrueType; Type 1 and CFF programs pass through untouched, so a font-heavy export from an older layout program usually needs rebuilding from its source instead. How to compress a PDF walks through the ordinary route first.

A fat Page content bucket surprises people who expected text to be cheap. Text is cheap; what costs here are vector drawings traced into thousands of tiny paths, and those streams already arrive compressed. If the PDF asks for a password before it opens, unlock it first: encrypted streams cannot be measured properly, and the report needs the plain object structure.

When the report says there is nothing to win

Sometimes the Next step panel admits no single part dominates and the file is already tightly distributed. Believe it: squeezing further only softens pictures. At that point the honest options sit elsewhere. Split PDF can carve off the chapters the recipient never asked for, Extract images can lift out the few pictures that matter, and the guide to compressing without losing quality explains what can still shrink with no visible change. Or skip the file surgery and share a cloud link instead.

Short answers about PDF file size

My breakdown says images are 90 percent. What now?
Press the recommendation button. It opens Compress PDF with the same file already loaded, so you can downsample the placed pictures without picking the file twice. Check the median placed resolution first: near 400 DPI means easy savings, near 150 means softer pictures.
Why do some table buckets not appear in my chart?
The bar only draws buckets that hold bytes. Attachments, metadata, or leftover objects at zero simply stay out of the picture instead of cluttering it.
Can compression shrink a font-heavy file?
Sometimes. Embedded TrueType gets subsetted to the glyphs the pages draw, but Type 1 and CFF programs pass through untouched. When those dominate, rebuild the document from its source with subsetting switched on.
The report calls my file tightly distributed. Is that good?
It means no single part dominates, so no single fix wins much. Either accept the size, split off unneeded pages, or share a link instead of the file.
Is the report really read-only?
Yes. It measures the object graph and prints numbers. No new PDF gets written, and the source file stays exactly as it loaded.

More tools that read before they write

The tools below work the same local way: the file stays with you, each run ends in a named download or a plain report, and none of them asks for an account.

Act on the breakdown