How browser-based image tools work without uploading files

The usual model: upload, process, download

Most online image tools send your file to a server, process it there and send the result back. For a while at least, your photo then exists on infrastructure you do not control, under someone else's retention, security and policies. Uploading also takes time, and large batches often run into size limits or queues.

The local model

Modern browsers can run compiled code at close to native speed with WebAssembly, and can do heavy work on background threads called Web Workers. Aback Images sends the image engine to your browser instead of sending your images to a server. Files are read from your disk into memory, decoded, processed and encoded inside the tab, and the result is saved back to your device.

What runs where

• Codecs: MozJPEG, OxiPNG, libwebp, libavif and libjxl compiled to WebAssembly, alongside the browser's own decoders where they are good enough.
• Other formats: libheif for HEIC in browsers without native support, pdf.js for PDF pages, and dedicated decoders for TIFF, PSD, camera RAW previews and more.
• AI tools: background removal, face detection and text recognition run small models with ONNX Runtime Web and Tesseract. Model and language files are downloaded from this site the first time and cached by your browser.
• Heavy codecs and models load only when a tool needs them, so opening a simple tool does not download all of them.

How to check it yourself

You do not have to take this on trust:

• Every tool shows an upload meter that counts bytes sent by the page. It should stay at 0 B while you work.
• Open your browser's developer tools, switch to the Network tab and watch while a tool runs. You will see downloads of app code and codecs, but no request carrying your image.

Trade-offs

Your device does the work, so very large images are limited by its memory. Images over roughly 100 megapixels may be refused to keep the tab from crashing, and phones reach their limits sooner. AVIF encoding, background removal and OCR are slower on an older phone than on a powerful server, and the first use of an AI tool downloads its model before it can run.

In exchange there is no upload wait, no account, no server-side file size cap and no copy of your images stored anywhere else.