Browser tool · Nothing installed

Reduce PDF Size Without Uploading It

Four quality levels and a target size, none behind a paid tier. The document is analysed and rewritten on this device.

PDF tooling by OnlinePCApps since 2013

Runs on device No sign-in Every level free Set a target size
0
Bytes leaving
<1s
Quality levels free
Documents per day
$0
Cost, permanently
Where a size limit applies

Three Limits That Force a Smaller PDF

Each is a fixed threshold imposed elsewhere rather than a defect in the document.

Attachment Limits

Gmail and Outlook reject attachments above roughly 25 MB. A scanned contract or a slide deck exported to PDF passes that figure without difficulty.

Reduces the document below the limit and reports the resulting size before it is saved.

Fixed Ceilings on Submission Portals

Government portals, job applications, visa forms and academic submissions routinely specify an exact figure: 100 KB, 200 KB, 500 KB or 1 MB, enforced by rejection rather than a warning.

Accepts an exact target and resolves the settings needed to reach it.

Storage and Distribution

Document archives and shared drives accumulate invoices and scans held at capture resolution rather than reading resolution.

Applies one resolution target across a document without altering its text or structure.
How it works

Three Steps to a Smaller PDF

instant PDF

Add the Document

Choose the document. The browser opens it from disk and examines its contents.

instant

Choose a Level or a Target

Pick a quality level, or state a target size and let the settings resolve towards it.

instant

Compare and Save

Original and compressed sizes appear side by side. Nothing is written to disk until the result is accepted.

Four quality levels Print, standard, screen and maximum, none paywalled
Target size State a figure and the settings resolve towards it
Before and after Both sizes shown before anything is saved
Compress a PDF Free, nothing to register The Desktop Build For directory-level runs
Against the category

Local Compression vs Server-Side Compressors

The established compressors transfer the document, process it remotely and apply a retention timer. Several also impose a daily task cap and reserve their stronger levels for a paid tier.

Property This page This page Runs on local hardware A typical server-side compressor Sends the document to a server
Compression levels at no cost All four Basic only, 20 to 40 percent
Target size setting Supported Rarely offered anywhere
Page count limit None 200 pages on some free tiers
Where the rewrite happens This machine A remote datacentre
Runs allowed per period Unmetered Two or three per day or hour
Largest file accepted Whatever memory allows 10 MB to 100 MB by provider
Copy held afterwards None exists Purged after one to twenty-four hours
Confidential material No restriction Subject to their policy
Effect of geography None Throughput falls with distance

That describes how the category generally operates and singles out no particular service.

Privacy

The PDF File Is Never Transmitted

Deletion timers are the standard reassurance here. One hour is typical, and the guarantee extends only as far as the operator's practices allow.

No timer applies where no document arrives. There is no endpoint capable of receiving a document and nothing of ours executes remotely. Open the browser developer tools, watch the network panel during a run and it stays empty throughout.

$ Network panel while compressing a 90 MB scan
external requests ....... 0
source document ....... 0 requests
0
Bytes leaving
0
Sign-ups
0
Copies kept
0
Third-party scripts
Compression mechanism

What Compression Removes and Rewrites

PDF compression is the process of rewriting a document so that it occupies less space: images are resampled and re-encoded, embedded fonts are reduced to the glyphs in use, and redundant objects are removed. It is usually presented as a slider with no account of what is being discarded. The table sets out each element, what happens to it and where the reduction actually comes from.

PartHeaderWhat it holds
Colour imagesdownsampled and re-encodedResolution is reduced to the chosen target in dots per inch, then the image is re-encoded as JPEG. This accounts for most of the reduction in a scanned document.
Greyscale imagesdownsampled and re-encodedHandled the same way as colour, at the same target resolution. Scans of printed text sit here rather than in the colour path.
Monochrome imagesdownsampled onlyBitonal scans are resampled but kept in a lossless CCITT or JBIG2 encoding, because JPEG artefacts are conspicuous on line art.
Image resolution300, 150, 96 or 72 dpiPrint quality keeps 300, standard drops to 150, screen uses 96 and maximum compression uses 72. Anything above the target is discarded.
Embedded fontssubsetOnly the glyphs the document actually uses are retained. A font embedded whole can carry thousands of unused characters.
Duplicate objectsmergedDeduplication means a logo repeated on forty pages is stored once and referenced forty times rather than embedded forty times.
Unused objectsdiscardedOrphaned dictionary entries left by earlier edits are removed and the cross-reference table is rebuilt. These accumulate quietly through repeated saves.
Page thumbnailsremovedCached preview images are regenerated by any reader on demand, so storing them adds size for nothing.
Document metadataretainedTitle, author and keyword fields are kept. They occupy a negligible amount of space and removing them serves no purpose.
Text contentuntouchedText lives in the content stream as drawing instructions rather than pixels and is already compact. It is never rasterised or degraded.
Vector artworkuntouchedLine art, charts and logos drawn as vectors stay vector. Converting them to images would increase size, not reduce it.
Form fieldsretainedAcroForm fields and their values survive compression, so a filled form stays fillable.
Alternatives

Which PDF Files Actually Shrink

Results vary by an order of magnitude depending on the input. The four cases below cover almost every document brought to a compressor.

Scanned Documents

A page scanned at 600 dpi carries roughly four times the data a 300 dpi page needs for the same legibility.
Reductions of 80 to 90 percent are ordinary here because the input was over-resolved to begin with.
Push the resolution too low and small print becomes unreadable. Check a page of body text before accepting the result.

Photographic Content

Photographs and screenshots re-encode well and account for most of the size in a presentation exported to PDF.
Typical reduction runs from 50 to 80 percent depending on the original encoding.
Repeated JPEG re-encoding compounds. Compressing an already compressed document degrades images for very little gain.

Text-Only Documents

A contract or report with no images is already close to its minimum size.
Expect single-digit percentages from font subsetting and object cleanup rather than a dramatic drop.
A tool promising 90 percent on a text-only PDF is either rasterising the pages or reporting a different file.

Already Optimised Files

A document compressed once has had its images downsampled and its objects cleaned already.
A second pass returns very little and costs image quality each time.
Where a file will not shrink further, the reason is that there is nothing left to remove, not that the tool failed.

Compress Before Merging

Note that one pass and smaller inputs produce a better result than compressing a combined file afterwards, because each source is optimised at its own resolution.

Compressing after a merge applies one setting to every source, including the pages that did not need it.

Encrypted and Signed Files

Note that a password and a signature both block the operation for different reasons: one prevents reading, the other cannot survive a rewrite.

Remove the restriction before compressing, and apply any signature to the compressed result rather than the original.
Desktop edition

Where PDF Compression Exceeds the Browser

Browser processing is bounded by two things: how much memory is available and the lack of direct disk access.

Where the Desktop Build Fits

Each page image is decoded, resampled and re-encoded in memory. A short document presents no difficulty. A scanned book of several hundred pages does, and a handset reaches that point long before a laptop.

Working page by page straight off the disk is what the desktop build does, and that removes the memory question.

Online viewerbounded by RAM
Desktop toolbounded by free disk
Never connects

Nothing reaches the network once installed, which suits controlled estates and offline workstations.

No memory ceiling

Pages are read from disk one at a time instead of being held together, so length stops mattering.

Folder-level batches

Apply one resolution target across a directory in a single pass.

Built for repeat work

Whole directories handled at a single resolution target, beyond what one tab can hold.

Saved profiles

Store a level and a target, then apply the same settings to future documents.

Runs disconnected

Requires no network access at all once set up.

About the Desktop Build Trial version · Windows 10 upward $39 outright · nothing recurring · refundable within 30 days
Reference

PDF Compression Questions

The OnlinePCApps PDF Compressor reduces the size of a PDF document entirely within the browser. Images are downsampled and re-encoded at a chosen resolution, embedded fonts are subset, duplicate objects are merged and unused entries are discarded. The file is read from local storage and written back to disk, with nothing uploaded and no level held behind a paid tier.
That depends almost entirely on what the document contains. Scanned pages commonly reduce by 80 to 90 percent because scanner output is captured at a far higher resolution than reading requires. Photographic content typically drops by 50 to 80 percent. A text-only report may reduce by only a few percent, because text is stored as drawing instructions and is already compact.
Because there was little to remove. Text and vector artwork occupy very little space, so a document containing only those is already near its minimum. A file that has been compressed once has also had its images downsampled already. In both cases the absence of a reduction indicates an efficient file rather than a failure.
Yes. All four levels and the target size setting are available without an account. Some providers reserve their stronger levels for a paid tier, which leaves the free option unable to reach the sizes people generally need.
Yes. State an exact figure and the settings resolve towards it. Common targets are 50 KB, 100 KB, 200 KB, 300 KB, 500 KB, 1 MB and 2 MB, because those are the values submission portals tend to specify. Most compressors offer fixed quality presets instead, which leaves the exact figure to trial and error.
Set 100 KB as the target and the resolution and encoding settings are chosen to reach it. Whether it is achievable depends on the content: a text-only document usually drops below 100 KB without difficulty, while a multi-page colour scan may not reach that figure at any resolution that keeps the text legible. The projected size is shown before anything is written.
100 KB and 200 KB are the most common for identity documents and application forms. 500 KB and 1 MB appear on academic and tender portals that accept multi-page submissions. 2 MB and above are typical of job application systems accepting a full document set.
No. Both the analysis and the rewrite occur inside the browser, on whichever machine holds the file. Nothing here can take delivery of a document. Open the browser tools and the network view stays silent for the whole run.
No. Text is stored as instructions referencing an embedded font rather than as pixels, so it is neither rasterised nor re-encoded. Text in a compressed document remains selectable, searchable and as sharp as the original.
Each font is subset so that only the glyphs the document uses are retained. A font embedded in full can carry several thousand characters for a document that uses eighty. Appearance is unchanged because the glyphs in use are the ones kept.
Print quality retains 300 dpi, standard reduces to 150, screen uses 96 and maximum compression uses 72. Images already below the target are left alone, because upsampling would add size without adding detail.
At print quality, yes, because 300 dpi is the figure commercial printing expects. Screen and maximum levels are intended for viewing and for email, and printing from them shows visible degradation in photographs.
Scans respond better than any other input. A page captured at 600 dpi holds roughly four times the data that 300 dpi requires for the same legibility, so the surplus is removed without affecting readability. Small print should be checked before accepting an aggressive level.
It can, but the second pass returns very little. Images have already been downsampled and re-encoded, and repeating JPEG encoding compounds its artefacts. Compressing once at the correct level produces a better result than compressing twice at a mild one.
Yes. Interactive fields and any values entered in them survive compression, so a filled form remains fillable and its data intact.
No. Compression rewrites the file, and a signature certifies one exact byte sequence. The field remains but no longer validates. Compress first and sign the result.
Only once the password is entered, since the content is unreadable without it. A file with owner restrictions may display normally yet still refuse the extraction step compression depends on.
No quota is applied. What binds the operation is available memory, a figure far higher than the ten to fifteen megabytes a server-side free tier typically allows.
No. There is no watermark, no daily allowance and no trial window. What comes out is the compressed document by itself.
Yes. Chrome, Firefox, Safari and Edge each run it, which puts macOS, Linux, ChromeOS, Android and iOS on the same footing as Windows. No installer is involved at any point.
A static page whose work happens on the visitor's own hardware costs almost nothing to host, and no document is ever received, processed or stored at this end. The desktop edition, bought by people running directory-level and oversized jobs, pays for the browser tools.
No. The source is opened read only and the compressed document is written as a separate file, so the original remains available if the result is not acceptable.
Authorship

Who Maintains This PDF Compressor

OnlinePCApps PDF Group

PDF internals since 2013

Compiled and checked by
13
Years at it
110
Utilities live
2
ISO standards used
0
Copies kept here

Document and file format tooling has been this company's work since 2013, starting with desktop software that read container formats directly rather than through a vendor library. The same approach runs here in JavaScript: decode each image stream, resample it to the chosen resolution, subset the fonts in use and rewrite the object table, all inside the browser process.

Standards referenced

ISO 32000 · PDF syntax JPEG · Image encoding Font subsetting FileReader API Image resampling

What it is tested against

Scans at 200, 300 and 600 dpi from flatbed hardware, sheet-fed units and phone camera capture Text-only documents, to confirm the tool reports a small reduction rather than inventing one Documents whose images are already below the target resolution Files already compressed once, to measure what a second pass actually returns Exports from PowerPoint and Excel that mix vector charts with photographs on one page Output opened in Acrobat, Preview and Chrome to confirm rendering is unchanged
Seen a document compress badly? Report it and the case gets looked at.

Reduce a PDF Without Transferring It

Nothing installed, nothing registered and nothing sent. All four quality levels cost nothing.

Compress a PDF Free, no account The Desktop Build For directory-level runs
Compress PDF · Free