PDF Compressor

PDF Compressor reduces file size with a two-track strategy: structural compression that rewrites the document with smaller object streams and stripped metadata, and rasterization that re-renders each page through a canvas and re-embeds it as JPEG at a lower DPI. The compressor decides which track to use by sampling the first five pages with pdf-lib and checking the ratio of XObject (image) resources to total page resources.

Track one — structural — runs unconditionally. pdf-lib serializes the output with object stream packing enabled, which collapses many small indirect objects into compressed streams. Optional metadata removal also clears the Info dictionary (Title, Author, Producer) and any XMP metadata. For text-heavy or vector-heavy PDFs at Low or Medium compression, this is the only step performed, so fonts and vector graphics are preserved at full fidelity.

Track two — rasterization — kicks in for image-heavy documents or whenever you select High compression. Each page is rendered through pdfjs-dist (Mozilla's PDF.js) onto an HTML canvas at a target DPI of 150 (Low), 96 (Medium), or 72 (High), then encoded as JPEG via canvas.toBlob with a quality factor of 0.85, 0.70, or 0.50. The result is a brand-new PDF where every page is a single JPEG image. Vector text becomes pixels in this mode — searchability and copy-paste are lost unless you run the result through OCR afterward.

Image-heavy scans benefit the most: a 40 MB scanned contract often compresses to 3–6 MB at Medium without visibly degrading legibility. Text-heavy PDFs (born-digital reports, invoices generated by software) usually shrink only 5–20 % from structural compression alone, because the text is already efficiently encoded.

There is also a target-size mode that iteratively re-runs rasterization with progressively lower DPI and JPEG quality until the output fits under the requested kilobytes. This is useful for portals with hard 5 MB or 10 MB upload caps. The search converges in 3–5 passes; on extreme targets it may stop at the lowest quality preset and inform you that the floor was reached.

The whole pipeline runs in your browser — pdf-lib for parsing and rewriting, pdfjs-dist (loaded as WebAssembly + worker) for rendering. No file leaves your device. Browser memory is the practical ceiling: PDFs up to ~150 MB compress comfortably on a modern laptop, but larger files may exhaust the tab's heap during canvas rendering.

Encrypted PDFs cannot be parsed by pdf-lib until they are unlocked; remove the password using the PDF Password tool first. If a digital signature is present, any modification — including compression — invalidates it; this is a property of the PDF signing standard, not a tool limitation.