All Offline Tools Directory

What Each Tool Has to Download Before It Can Run

Everything here runs locally, but “locally” still has a cost, and it is not the same cost for every tool. Some do their work with code that arrived with the page. Others need a compiled engine fetched on first use. Knowing which is which explains the one thing people find surprising: why merging a PDF is instant and the first video conversion makes you wait.

Nothing extra — it works the moment the page loads

The text and data utilities (Base64, URL encoding, JSON and CSV handling, the mock data generator), the SVG optimiser, the EXIF stripper, and the whole family of PDF page operations fall in this group. They are plain JavaScript working on bytes already in memory, or they lean on browser APIs — canvas, ImageDecoder, createImageBitmap — that your browser ships with. Image conversion mostly lives here too: the JPEG, PNG and WebP encoders are part of the browser, so converting between them is a decode and a re-encode with no download at all.

A small module, fetched once

A few jobs have no browser API behind them and need real compiled code. Extracting a ZIP, 7z, TAR or RAR archive pulls in a WebAssembly build of libarchive; removing a PDF password pulls in qpdf. Each is a single WebAssembly file of about a megabyte, cached by your browser afterwards, and it is fetched when you open that tool rather than on every page of the site.

A large one-time download

Three things are genuinely heavy. Video and audio conversion runs FFmpeg compiled to WebAssembly: about 10 MB gzipped over the wire, unpacking to roughly 32 MB of code that then lives in your browser’s cache, so the second conversion starts immediately and the first one does not. OCR downloads a trained language model before it can read anything, and background removal downloads a segmentation model. There is no way around this for the same reason the tools are private at all: a server-based converter keeps the codec on the server and sends you only the result, so the price of never uploading your file is downloading the machinery instead. The trade is a slow first run and then an unlimited, offline-capable tool, rather than a fast first run and a queue on someone else’s hardware.

All of it comes from us

Most browser-based converters load these engines from a public CDN, which quietly undoes part of the point: the CDN cannot see your file, but it can see your IP address and which tool you just opened. Every asset above is served from this site or from our ownassets.duckconvert.com storage, so opening the OCR tool tells a third party nothing, because no third party is asked. Once an engine is cached, most tools here keep working with no network connection at all — a reasonable test of whether a converter really runs where it claims to.