Merge PDF
Combine PDFs and reorder pages by dragging. Nothing is uploaded.
Re-encode embedded images to cut file size without a server.
Almost always images. A page of text is a few kilobytes; a scanned page at 300 DPI in full color is one to three megabytes before compression.
| What is inside | Typical size per page | Compressible |
|---|---|---|
| Text with embedded fonts | 5–50 KB | Barely — already efficient |
| Vector graphics and charts | 20–200 KB | Somewhat |
| Photograph at web quality | 100–400 KB | Yes |
| 300 DPI color scan | 1–3 MB | Substantially |
| 600 DPI color scan | 4–12 MB | Dramatically |
| Unsubsetted embedded font | 200 KB–2 MB each | Yes, by subsetting |
The second most common cause is font embedding without subsetting. A full CJK font can be several megabytes, and a document that embeds four of them in full carries that weight on every copy. Subsetting keeps only the glyphs actually used and usually recovers 90% of it.
Each level is a resolution target plus a JPEG quality for the raster content. The document structure, text layer and vector content are left alone.
| Level | Image DPI | Quality | Typical reduction | Use for |
|---|---|---|---|---|
| Light | 200 | 0.90 | 30–50% | Print-quality archives |
| Balanced | 150 | 0.80 | 60–80% | Email, general sharing |
| Strong | 120 | 0.70 | 80–90% | Web upload, form submission |
| Maximum | 96 | 0.60 | 90–95% | Screen viewing only |
Balanced is the right default for the common case of emailing a scanned document. Most corporate mail servers reject attachments over 10–25 MB, and Gmail’s limit is 25 MB; Balanced brings almost any real-world scan comfortably under that.
That depends on whether the text is real text or a picture of text.
In a digital PDF exported from Word or InDesign, text is stored as font references and vector outlines. It is resolution-independent, compression does not touch it, and it stays perfectly sharp at any zoom.
In a scanned PDF there is no text — only an image of a page. Compressing that image compresses the text along with it, and at aggressive settings the characters develop visible artifacts. For scans, stay at Balanced or Light, and consider running the document through OCR first so the text becomes selectable and the image quality matters less.
Because there was little raster content to compress. A 40-page text document is already close to its floor: fonts and vector outlines are compact by nature, and there is nothing to re-encode.
Three other cases produce the same result. A PDF whose images are already heavily compressed cannot be compressed much further without visible damage. A PDF containing mostly vector charts has no raster layer to target. And a PDF that is large because of embedded attachments or an enormous form structure needs those removed rather than re-encoded.
The tool reports the composition of the file before you compress, so you can see whether images are actually the problem.
Scanned documents typically shrink by 80–95%. Text-only documents shrink by 5–20% at most, because there is very little to compress.
No. The text layer is preserved. Only embedded images are re-encoded.
Set a target size and the tool iterates the quality setting until the output fits, or reports the closest it can reach without destroying legibility.
No. Re-encoding images at lower quality discards data permanently. Keep your original.
Usually a size cap, commonly 2 MB or 5 MB. Use Strong or Maximum, or split the document and submit it in parts.
Only if you supply the password, since the content must be decrypted before it can be re-encoded.
No. Rendering and re-encoding both use your own CPU and GPU inside this tab.