How in-browser file tools work: JavaScript, WebAssembly and on-device AI
Ten years ago, converting a file online almost always meant uploading it. Today a web page can do much of that work on your own device. Here is what changed, in plain language.
- Published
- CategoryAI & Technology
- Reading time3 min read

The browser became a capable computer
A web page is no longer just text and pictures. Modern browsers include fast JavaScript engines, graphics and audio interfaces, and ways to read files you select and save results back as downloads. Together they let a page behave like a small desktop application while it is open.
Crucially, reading a file you pick is not the same as uploading it. When you choose a file, the browser gives the page access to that file's contents in memory. Unless the page's code sends the data over the network, it stays on your device.
The building blocks
JavaScript and the File API
JavaScript runs the page's logic. The File API lets it read the bytes of a file you chose; the Blob and download mechanisms let it hand back a new file. Simple tools such as text formatters, word counters, unit converters and QR code generators need little more.
Canvas and image decoding
Browsers already know how to decode JPG, PNG, WebP and other formats in order to display them. Through the canvas interface, a page can draw an image, resize or crop it, and encode it again as a new JPG, PNG or WebP. That is how in-browser image resizers and compressors work.
Web Audio
The Web Audio interface decodes audio files your browser can play into raw samples. A page can then cut, change speed or analyse the sound and encode a new file, for example with an MP3 encoder written in JavaScript. The Convertios MP3 Cutter works this way.
WebAssembly
WebAssembly (often shortened to Wasm) is a compact binary format that browsers run at near-native speed. Developers can compile existing software written in languages such as C, C++ or Rust into WebAssembly and use it inside a page. This brings heavyweight libraries, for example for PDF parsing or image processing, into the browser without rewriting them.
On-device AI with WebGPU and WebAssembly
Machine-learning models can also run in the browser. A model file is downloaded once, often cached for later visits, and inference runs with WebAssembly or, where available, the GPU through WebGPU or WebGL. The Convertios Remove Background tool uses the open-source IMG.LY background-removal library this way: on first use the browser downloads and caches about 40 MB of model and runtime files, then processes your image locally. See how AI background removal works.
What this means for you
- Privacy: files processed locally never reach a server. For PDFs, Merge PDF combines documents in the browser using a PDF library.
- Speed: there is no upload or download wait for the file itself; work starts immediately.
- Offline use: some tools keep working after the page has loaded, even without a connection.
The limits
- Memory: everything happens in your device's memory. Very large videos, hour-long recordings or huge PDFs can be slow or fail on older phones.
- First-load downloads: tools with large libraries or AI models must download them first, which costs data on mobile.
- Browser differences: a format your browser cannot decode, for example some HEIC photos in some browsers, cannot be opened by a page that relies on the browser's decoder.
- Some tasks still need servers: transcoding video with FFmpeg or rebuilding a PDF as Word are handled on the Convertios backend because they need more resources. Our comparison of browser and server processing lists which Convertios tools work which way.
How to tell whether a tool is local
Look for statements such as "processed in your browser" on the tool page, check the browser's network panel for large uploads, or disconnect from the internet after the page loads and try again.
Published by Convertios.


