Compress PDF to Any Size — Text Stays Sharp & Selectable, Free

Upload forms cap PDFs at 100 KB, 200 KB or 500 KB, and the files people actually have are far bigger: a one- or two-page document saved from Word can weigh 250 KB or more, and a phone scan often passes a megabyte. Tools that promise an exact size often get there by turning every page into a JPEG picture and lowering its quality until it fits. It works, but the text stops being text: it can no longer be selected or searched, and at small sizes it turns blurry.

This compressor works in three steps and stops as soon as the file fits. First, nothing you can see changes. Data the document no longer uses is removed, identical copies are merged, and each embedded font is trimmed to the letters the document actually uses. Word often embeds large parts of Calibri or Arial Unicode MS, sometimes the same font several times, and in office PDFs that is usually most of the file. Every page is then drawn again and compared with the original; if any page changes, that font is put back. Second, only the pictures inside the PDF are recompressed: lower resolution, a more efficient JPEG encoder, and true grayscale for grayscale scans. Text, fonts and layout are not touched. Only if neither is enough do the pages become images, and the result tells you so.

We measured it on 23 real PDFs published by Indian government offices. The first step alone made 14 of them between 12% and 86% smaller: 4.5 MB became 1.3 MB, and 292 KB became 75 KB. In all 14, every page looked identical when drawn, with the same text, in three PDF engines: PDFium (used by Chrome and Edge), pdf.js (Firefox) and MuPDF. In the other 9 the fonts could not be trimmed (already small, or of an older type), and those came out the same size or up to 11% smaller. We also compared it with the page-as-image method on test documents we built from one of those reports, with stock photos added and scanning simulated. A three-page document with photos fit in 100 KB with all its text selectable and 99.7% of its words readable by OCR, against 71% as page images. A three-page grayscale scan at 100 KB came out at 87% against 60%.

The honest limits. PDFs with a password or with security restrictions are left alone, because compressing them would save them without that protection. A digitally signed PDF can be compressed, but any change to the file makes the signature show as invalid; the tool warns you, and if the signature matters, upload the original. In fillable forms the fonts are kept whole, because whatever someone types later needs every letter. When the target is smaller than the text and fonts alone, the pages do become images. The result then says "pages as images", and the text, links, bookmarks and form fields are not kept; at very small sizes, check that small text is still readable. Pictures in print colours (CMYK) and black-and-white fax-style scans are not recompressed yet, so PDFs made of them reach small targets only as page images. Long documents take longer: a 141-page manual took between 10 and 45 seconds on a laptop. Everything runs in your browser, so the PDF is never uploaded.