Methodology
Last updated: April 15, 2026
Where processing happens
FileCurve runs the majority of its tools inside your browser using a mix of Canvas 2D, WebAssembly, and the Web Audio API. When you upload a file to an image compressor, a PDF merger, or a meme generator, the bytes stay on your device. The page reads the file through the File System Access API or a standard input, performs the requested operation on a Web Worker where practical, and returns a download.
Libraries we use
PDF operations such as merge, split, rotate, watermark, and page-number insertion are powered by pdf-lib, a pure JavaScript PDF library that reads and writes standard PDF 1.7 documents without a native dependency. Image compression uses browser-image-compression for JPEG and WebP, with a direct Canvas fallback for PNG and BMP. Raster resizing uses createImageBitmap with resize quality hints for sharper output than a plain 2D context draw.
When we use a server
Some tasks cannot run in the browser at a reasonable speed. Long video transcoding, OCR on large documents, and background removal on high-resolution images run inside short-lived Vercel functions. These functions receive the file over HTTPS, process it in memory or in a temporary directory, return the result, and then the input and output are scheduled for deletion within one hour. No file is retained beyond that window, and logs do not include file contents.
Quality and defaults
Every tool ships with a default that targets a sensible quality-to-size ratio. Image compression defaults to a quality setting that keeps the output perceptually close to the source while producing a visibly smaller file. PDF compression rasterises images at 150 DPI for the standard preset and 96 DPI for the aggressive preset. Advanced users can override these values on the tool page before processing.
Updates and corrections
When a format spec changes, a library ships a breaking release, or a reader points out a better default, we update the relevant tool and note the change in the changelog. Contact hello@filecurve.com with corrections. The author is Aniket Nigam.