Compress PDF Files Online

1. 1

   Add your PDF

   Drop the file in or browse for it. The tool reads the document straight away and shows you where its bytes are — how much is pictures, how much is fonts, how much is the page description itself.

2. 2

   Pick a mode

   Lossless rewrites the structure and changes no pixel. Balanced and Strong recompress the pictures inside the file and leave text, links and form fields untouched. Flatten turns pages into images. Each mode shows its estimated output size before you run it.

3. 3

   Compress

   Press Compress. Progress is shown per image, and you can cancel at any point. Nothing is uploaded — the file is read and rewritten on your own device.

4. 4

   Check the result, then download

   The result card shows the new size, whether the text survived, and a page-quality score measured against the original. Compare the before and after thumbnails, then download.

A PDF that will not send is a PDF that has stopped being useful. Mail servers reject attachments over 10 or 25 MB, upload forms cap out at 2 MB, and a 40 MB scan takes long enough to open on a phone that people give up on it. Compressing a PDF fixes all three — but only if the compressed file is still the document you started with. This tool reduces file size while keeping text selectable, links clickable and form fields fillable, and it measures what it did to every page so you are never guessing.

Benefits of PDF Compression

• Fits the limit: name a target size and the tool searches for settings that reach it
• Still a document: text, links, bookmarks and form fields survive in three of the four modes
• Informed, not guessed: see where the file's bytes are before choosing how to attack them
• Proven, not promised: a page-quality score measured against the original after every run
• Private by construction: the file is never uploaded, so there is nothing to delete afterwards

There are two completely different ways to make a PDF smaller, and most tools do not tell you which one you just used. The destructive way is to draw each page as a picture and throw away the page description — the file shrinks dramatically, and the document stops being a document: you cannot select a sentence, search for a word, click a link, fill a field or copy a figure out of a table. The careful way is to leave the page description exactly as it is and re-encode only the pictures embedded in it. That is what Lossless, Balanced and Strong all do here.

What survives, and how we know

• Every character of text: after each run the tool extracts the text from the result and compares it with the source. The result card tells you whether they matched.
• Every hyperlink: link annotations are counted before and after and both numbers are reported.
• Every form field: form widgets are counted the same way, so a filled-in form stays fillable.
• Bookmarks and the outline: the document catalogue is rewritten, never rebuilt, so the navigation tree comes through intact.
• Vector drawings and tables: charts stay sharp at any zoom because they were never turned into pixels in the first place.

The one mode that does not do this is Flatten, and it is labelled as the mode that turns pages into images. Choosing it is a decision, not an accident — which is the whole difference between this tool and a three-button compressor that quietly rasterises your contract at its middle setting.

Before you pick a compression level, it is worth knowing what you are compressing. A PDF is not one thing: it is a page description, a set of embedded fonts, a pile of image streams and a cross-reference table, and the proportions differ wildly between documents. The tool reads all of this the moment a file is added and shows it as a byte budget, because the right answer for a 30 MB scan and a 30 MB architectural drawing are not the same answer.

The five categories

• Images: every raster picture in the document. On a scan this is 95 per cent or more of the file, and it is the only category a lossy mode can meaningfully reduce.
• Fonts: embedded typefaces. A document with many fonts subset badly can carry several megabytes here, and no image compressor will touch it. Where the same typeface has been embedded more than once — which merging documents always causes — the copies are merged into one.
• Page content: the drawing instructions — text runs, vector paths, tables. Lossless rewriting compacts these; nothing else changes them.
• Metadata: the title, author, editing history and application bookkeeping. Small, but often the part people most want removed before sharing.
• Structure: the cross-reference table, object headers and the trailer. Rewriting the file with modern object streams reclaims most of the waste here.

The panel also names the largest single image in the document and the density it is drawn at, which is usually the whole story: a 4 000-pixel-wide photograph placed into a 6-inch-wide column is stored at roughly 670 dots per inch, several times more than any screen or office printer can use.

Most size problems are not "make this smaller", they are "make this fit under 5 MB". Guessing at a compression level and re-running until something fits is the usual answer and it is a poor one. Enter the size you need instead and the tool searches for it directly, halving in on the gentlest settings that come in under the limit.

How the search works

• Bounded: at most six passes, always. A search with no stated bound is a progress bar that might never end.
• Seeded: the first guess comes from the document's own byte budget, which usually saves two of the six.
• Gentlest wins: when several settings fit, the search keeps looking for a softer one rather than stopping at the first that works.
• Honest when it cannot: if the target is unreachable even at maximum strength, the tool says so and gives you the smallest file it could make, with that size named.
• Never re-compressed: every pass starts from the original pictures, so the sixth attempt is not a compression of the fifth.

Flatten draws every page as a single picture at a chosen density and builds a new document from those pictures. It is the smallest result and the most destructive one, and there are two good reasons to choose it deliberately.

Good reasons, and one bad one

• You want the text gone: a flattened page cannot be selected, copied or searched, which is occasionally exactly the point when circulating a draft.
• Appearance must be frozen: unusual fonts, transparency groups and overprint all render identically everywhere once the page is a picture.
• Not for size alone: on most documents Strong produces a smaller file than Flatten while keeping the text, because Flatten also has to encode all the white space around the words.

Flattened pages keep their original paper size, so a flattened A4 document still prints as A4 rather than as a picture measured in pixels.

Compression happens in two independent places: the file's structure, and the pictures inside it. Every mode does the first. Three of the four do the second, at different strengths, and only Flatten replaces the pages themselves.

Compression Methods

• Lossless: structure and object streams only. Not one pixel changes, and the rendered page is identical.
• Balanced: pictures downsampled to 150 dots per inch and re-encoded. The recommendation for almost everything.
• Strong: pictures downsampled to 96 dots per inch at lower quality. For attachment limits.
• Flatten: pages redrawn as images. Smallest, and the only mode that removes text.
• Fine controls: quality, density, grayscale and metadata removal, for when a preset is close but not quite right.

A few things that are true of PDFs generally, and worth knowing before you compress one.

Key Statistics

• 150 dots per inch is the practical ceiling for on-screen reading and ordinary office printing; most scans arrive at 300 or 600
• A page stored at 600 dpi carries four times as many samples as the same page at 300 dpi, for no visible gain on screen
• Re-encoding a picture that is already well compressed can make it larger — the tool checks, and keeps the original when that happens
• Bilevel scans stored as CCITT or JBIG2 already beat anything a photographic codec can do, so they are left alone
• JPEG has no transparency channel, so any transparent region has to be composited onto something before re-encoding
• Encrypted PDFs must be unlocked before any tool can rewrite them

How to get a small file that is still worth having.

Quality Considerations

• Start with Balanced and only go to Strong if the size still is not right
• Read the byte budget first — if the file is mostly fonts, no image mode will help much
• Check the page-quality score and the worst-page figure, not just the percentage saved
• Compare the before and after thumbnails on any document with fine print or thin lines
• Use the target-size field rather than guessing at a level when you have a hard limit
• Keep Flatten for the cases where losing the text is the point

Common Use Cases

• Email attachments: set a target of 9 MB and let the search find the gentlest settings that fit
• Scanned paperwork: Strong, where a 600 dpi scan drops to a readable 96 dpi and loses most of its size
• Portfolios and reports: Balanced, so the photographs shrink while the captions stay crisp
• Contracts and forms: Balanced or Lossless, because the fields have to stay fillable
• Website downloads: Balanced with metadata removed, so nothing about the author ships with the file

Under the hood, compressing a PDF without destroying it is a matter of touching only the streams that hold pixels and leaving every other object in the file exactly where it was.

Compression Process

• Read the object graph: every indirect object in the file is walked and priced into a category.
• Find the pictures: image streams are identified along with their colour space, bit depth and the codec they use.
• Work out the real density: each page's drawing instructions are scanned to find how large each picture is actually drawn, which is what turns pixels into dots per inch.
• Downsample and re-encode: only pictures with more samples than the target density needs, and only where the replacement is genuinely smaller.
• Rewrite, do not rebuild: the new stream is written back into the same object, so everything that referenced it still does.
• Measure the result: both documents are drawn at the same density and compared, and the text of each is extracted and checked.

Everything runs on your device — nothing is uploaded, and there is no queue to wait in.

• PDF merger
• PDF splitter
• PDF watermark
• Image to PDF
• Compress PDF to 100KB
• Compress PDF to 200KB