Simon Willison released Photo Scrubber on 29 September 2026 to address his discomfort with sharing images containing identifiable faces of strangers. The browser-based tool runs entirely locally, meaning no photographs leave the user’s device during processing. It combines Google’s MediaPipe C++ library compiled to WebAssembly with the BlazeFace detection model to automatically identify and blur human faces. Users can adjust blur strength or perform manual redactions before exporting the final image. The application strips all personal metadata from the exported files, ensuring that embedded data such as GPS coordinates or camera settings do not accompany the shared picture. This approach prioritises privacy by keeping sensitive visual data on the client side rather than sending it to a remote server for analysis.
The utility matters because it provides a practical solution for photographers who wish to protect individual privacy without relying on external cloud services. By executing complex machine learning tasks within the browser, the tool eliminates the risk of data exposure inherent in online editing platforms. It also removes the need for manual cropping or editing to obscure identities, streamlining the workflow for ethical image sharing.
- Runs entirely in the browser without server uploads
- Uses MediaPipe and BlazeFace for face detection
- Strips EXIF and other metadata from exports
