Simon Willison releases Photo Scrubber, a local face-blurring tool

On September 30, 2026, Simon Willison released Photo Scrubber, an experimental tool that detects and blurs faces in photographs and removes their metadata, with the processing carried out locally.

Willison said the idea came from a concrete experience: he took a photograph of protesters and then found he did not want to share pictures of strangers whose faces were identifiable. He had GPT-6 Astra build the tool for him, with the aim of automatically identifying faces and blurring them out.

On the technical side, Photo Scrubber uses Google's MediaPipe C++ library, compiled to WebAssembly via @mediapipe/tasks-vision, together with the BlazeFace model for face detection. Blurring is applied automatically to the faces it finds, and the tool additionally strips metadata from the image files it processes. Willison frames both capabilities as privacy protection, and the work happens on the user's own machine rather than through a remote service.

He listed the project in his site's tool section under the heading "Tool: Photo Scrubber — local face blur & metadata removal," describing it as a local face blur and metadata removal utility.

That is as far as the available material goes. The announcement is the only report on the tool so far, and it does not say which platforms Photo Scrubber runs on, how it is distributed, or what changed after the initial release. It also does not address accuracy, performance, or limitations of the face detection. The post presents the project as experimental and credits GPT-6 Astra with building it.