Skip to content

PDF size analyzer

See what makes a PDF so large.

Drop a PDF here

Only your browser reads them.

How to use Size analyzer

  1. Open your PDF. Drop a file onto the analyzer or click Choose a file to start scanning immediately.
  2. Inspect the category bar. Review how many bytes and what share go to Images, Fonts, Page text and drawings, Attached files, and Structure and other.
  3. Check the Largest images table. Note the pixel dimensions, compression format, and page locations of heavy graphics.
  4. Review Largest fonts and advice. Read the sidebar recommendation to see if image compression or font pruning is needed.
  5. Act on the findings. Click Compress this PDF to send the file directly into the compressor without re-uploading.

Your files are processed on this device and never uploaded. They stay private, and there's nothing to delete afterwards.

How does the PDF size analyzer measure stream objects?

The analyzer enumerates every indirect object inside your document and sums exact stored byte lengths across five categories: Images, Fonts, Page text and drawings, Attached files, and Structure and other. As soon as you drop a file, a background worker parses the object tree and renders a proportional bar alongside exact byte counts.

Counting exact bytes across five PDF categories

Under the Library of Congress PDF 1.7 (ISO 32000-1) Format Description, a PDF packages visual assets and page instructions as numbered indirect streams. The worker classifies Image XObjects, FontDescriptor font programs, page Contents and Form XObjects, and EmbeddedFile attachments into separate buckets.

Any bytes outside those streams, such as page tree dictionaries and cross-reference tables, flow into Structure and other. If a damaged cross-reference table blocks inspection, you can fix corrupted PDF cross-reference tables first before re-running the audit.

Excluding transparency masks and object stream overhead

To avoid inflating image counts, the worker skips helper streams tagged as ObjStm or XRef when tallying individual buckets. It also checks SMask, Mask, and SMaskInData references so alpha transparency channels are not listed twice in Largest images.

Recursive traversal through nested Form XObjects tracks exactly which 1-based page numbers draw each picture. That way, you can spot whether a heavy logo repeats across 50 pages or sits on page 1 alone.

Auditing the largest images, filters, and embedded fonts

Below the category summary, the report lists up to the 10 largest images and 10 largest embedded fonts sorted from heaviest to lightest. Each image row displays its raw pixel width and height, its decoded filter label, and the page numbers where it appears.

Decoding image stream filters and pixel dimensions

PDF stream filter Displayed format label Typical content type Compression behavior
DCTDecode JPEG Photographs and scans Lossy continuous-tone
JPXDecode JPEG 2000 High-bit-depth scans Wavelet lossy or lossless
FlateDecode Compressed Screenshots and vectors Lossless Deflate
CCITTFaxDecode Fax (black and white) Legacy 1-bit scans Lossless bi-level
JBIG2Decode JBIG2 (black and white) Modern 1-bit scans Pattern-clustered bi-level

When a photographic scan uses DCTDecode as described in the Library of Congress JPEG Image Encoding Description, excessive pixel dimensions such as 3500 × 4900 px usually explain the bloat. Conversely, uncompressed or lossless FlateDecode screenshots often shrink dramatically when converted or downsampled.

If your audit reveals heavy form fields, overlapping annotations, or layered vector transparencies bloating page content, run a PDF flattening tool to merge interactive layers into compact static streams.

Spotting embedded font bloat and subset prefixes

The Largest fonts list inspects FontFile, FontFile2, and FontFile3 streams linked to each FontDescriptor, stripping standard 6-letter subset prefixes to show clean typeface names. Multi-language or unsubsetted desktop fonts can easily add 500 KB to 3 MB per typeface.

Comparing client-side analysis versus cloud audit services highlights a major speed advantage: local inspection reads raw byte buffers immediately without waiting for multi-megabyte uploads over slow connections. You also avoid leaving confidential client statements on external servers just to inspect why a file grew.

Why do some bloated PDFs resist standard compression?

A PDF resists standard compression whenever its largest bytes sit inside already-compressed binary streams rather than oversized bitmaps. Font programs, ZIP-compressed attachments, and page content streams encoded with FlateDecode follow the IETF RFC 1951 DEFLATE Compressed Data Format, leaving minimal redundancy for lossless re-compression.

Dominant category Typical share Does image compression help? Recommended next step
Images 50% or more Yes, substantial savings Downsample via Compress PDF
Fonts 40% or more Little to no reduction Re-export with font subsetting
Attached files Varies No effect on attachments Extract and remove attachments
Page text and drawings High on CAD files Minimal reduction Flatten complex vector layers

Checking the breakdown before compressing saves time and prevents quality loss. When Images dominate, click Compress this PDF in the sidebar to shrink heavy image streams using Medium or Strong settings without re-uploading your file.

Turning audit findings into smaller, private documents

Every byte calculation happens in browser memory, sending 0 bytes of your PDF across the network. Auditing invoices, medical scans, and legal discovery files locally keeps sensitive metadata private while helping you diagnose large PDF files for email attachment limits.

Acting on the 50% image and 40% font thresholds

The right-hand panel evaluates your top category automatically. When images reach 50 percent or more of the file, it recommends Medium or Strong compression; when fonts reach 40 percent or more, it warns that image compression will barely move the needle.

Tip: Files smaller than 500 KB (500,000 bytes) are flagged as already small, so you can skip aggressive re-encoding that might blur fine text.

When a multi-chapter report remains above a 25 MB email attachment limit even after image compression, you can split oversized documents for email limits into smaller standalone parts.

Auditing sensitive files locally in a Web Worker

As documented in the MDN Web Workers API Documentation, running object-tree enumeration in a dedicated worker thread prevents browser freezes on 100-page files. Clicking Compress this PDF transfers the in-memory file directly into the compressor tab for immediate action without re-selecting it from disk.

Questions

How exact are the byte numbers in the size report?

Stream sizes for images, fonts, page content, and embedded attachments are measured directly from their stored byte lengths inside the PDF. The Structure and other row accounts for dictionaries, cross-reference tables, and remaining container bytes.

What takes up the most space in a PDF?

High-resolution embedded images are the most common cause of oversized PDFs, followed by unsubsetted fonts, embedded file attachments, and dense vector drawings. The stacked bar at the top of the report shows the exact percentage for each category.

Why does font bloat resist standard PDF compression?

Embedded TrueType and OpenType font programs are binary files that are often already Flate-compressed inside the PDF. Downsampling image streams cannot shrink font tables, so font-heavy PDFs stay roughly the same size after standard image compression.

What are Attached files inside a PDF size breakdown?

PDF containers can store complete EmbeddedFile streams inside the document, such as source spreadsheets, XML invoices, or raw CAD files. These attachments add to file size even though they never appear visually on any page.

Are transparency masks counted as separate images?

No. Soft masks and alpha mask streams referenced by SMask or Mask entries are excluded from the standalone Largest images list so transparent graphics are not double-counted as separate pictures.

What happens if my PDF fails to parse during analysis?

Severely damaged cross-reference tables or truncated streams can prevent the parser from enumerating indirect objects. Running the file through Repair PDF first rebuilds broken structure tables so the analyzer can inspect every stream.

Updated