Most PDF tools read your entire 500 MB file to change a single page. We read 4 KB.
The engine reads the last ~4 KB of your PDF structural index. Your text, images, and content are never touched.
Synapse constructs a surgical delta new objects, fonts, xref table that appends after the original file.
Your browser concatenates original + patch. The original bytes are never modified. Validated by qpdf + mutool.
Traditional tools upload 100% of your file. We read less than 0.01% — only the structural metadata where the page index lives. The engine is physically incapable of accessing your content.
The Synapse engine never reads your text, images, or form data. It only reads the PDF structural tail — byte offsets and page pointers.
Everything runs in your browser’s JavaScript engine. There is no API, no cloud function, no upload endpoint.
Your original bytes are never modified. The engine appends a surgical patch after %%EOF. The file is never re-serialized.
OnlyPDF is a static site. No databases. No accounts. No logs. Your documents exist only in your browser’s memory and vanish the moment you close the tab.