A browser-only tool sends you the program instead of asking for your file. When you open the page, your browser downloads the tool’s code; when you add a file, the browser reads it from your disk and processes it on your own computer; when you click Download, the result goes straight from your browser’s memory to your device. The website’s server never receives the file, so it can’t keep, leak or look at it.
Two ways a file tool can work
| Upload-based tool | Browser-only tool | |
|---|---|---|
| Where the work happens | The company’s server | Your computer or phone |
| Your file travels | To the server and back | Nowhere |
| Speed depends on | Your upload speed and their queue | Your device |
| Size limits | Often capped on free plans | Your device’s memory |
| Works offline after loading | No | Often, for much of the work |
| Can save your work for you | Yes, if you have an account | No: you download it |
Most well-known online file tools are upload-based; we looked at what six of them say about it in Which online file tools upload your files?
What makes it possible
Ten years ago, a web page couldn’t realistically edit a PDF or convert a video. Several things changed that:
- WebAssembly. A way to run fast, compiled programs inside a web page, at close to the speed of a desktop app. It became an official W3C web standard on 5 December 2019, described as “a virtual instruction set architecture that enables high-performance applications on the Web” (W3C). It lets existing software, written in languages like C, C++ and Rust, run in the browser.
- Access to your device’s media hardware. The WebCodecs interface lets a page use the same chips your device uses to record and play video, so converting a video in a browser can be as fast as in a desktop app.
- Access to your graphics card. WebGPU lets a page run heavy maths, such as an AI model, on the graphics processor.
- Background threads. Web Workers let a page do heavy work without freezing, so you can keep using it while a batch of 300 photos processes.
- Reading and saving files locally. Browsers can read a file you choose and save a result, all without a server in between. Chrome and Edge can even write results straight into a folder you pick.
How each JFDI tool uses them
| Tool | What runs in your browser |
|---|---|
| PDF toolkit | Mozilla’s PDF.js draws the pages; pdf-lib rebuilds the file with your changes |
| File converter | SheetJS for spreadsheets; Mediabunny with WebCodecs for video and audio |
| Image editor | A HEIC decoder compiled to WebAssembly; the IS-Net background-removal model on WebGPU, or WebAssembly when there’s no usable graphics card |
| Chart maker and map maker | The d3 drawing library, plus boundary files served from this site |
| QR code maker | A QR encoder written in plain JavaScript |
All of these are open-source libraries, credited in our licences file.
What still gets downloaded
“Nothing is uploaded” doesn’t mean “nothing is downloaded”. Your browser fetches:
- The tool’s code, when you open the page.
- Extra pieces when they’re needed: the PDF engines in the background after the PDF toolkit opens, the 44 MB background-removal model the first time you use it, map boundaries for the US or UK when you make those maps. Your browser keeps them, so later visits are quick.
- The font for the pages, from Google Fonts. Like any website, Google can see your IP address when your browser fetches it. Our privacy page explains this.
None of these requests contain your files.
How to check for yourself
Open the browser’s developer tools (F12), choose the Network tab, and use the tool. You’ll see every request the page makes. An upload shows up as an outgoing request about the size of your file; on a browser-only tool, there isn’t one. The steps are in Which online file tools upload your files?
The trade-offs
- Your device does the work, so a huge video converts faster on a laptop than an old phone.
- Nothing is saved for you. That’s the privacy, but it also means closing the tab loses unsaved work. Download as you go.
- Some jobs still need a server, such as recognising text in scanned documents (OCR) or reading formats browsers can’t, like old Word
.docfiles.