In the age of ubiquitous facial recognition, uploading a video of a protest, a strike, or even a party can inadvertently put your friends and allies at risk. The surveillance state scrapes public platforms, building databases of faces linked to locations and events. If you are documenting resistance, you have a duty of care to the people in your frame. Blurring faces manually in Premiere or DaVinci Resolve is tedious and time-consuming. This is where Deface comes in. What is Deface? Deface is an open-source command-line tool that uses machine learning to automatically detect and anonymize faces in videos and photos. It's fast, local, and doesn't send your footage to the cloud. "By default all audio tracks are discarded as well." — This is a feature, not a bug. Voices can be biometric identifiers too. Installation You'll need Python installed. If you're on a mac/linux/windows machine with a terminal: Basic Usage The simplest way to use it is to just point it at a video file: This will output with all detected faces blurred. Advanced Tactics Adjusting Thresholds The flag controls how sensitive the face detection is. The default is . Lower (e.g., 0.1): Detects more faces, but might mistake a mailbox for a person. Higher (e.g., 0.5): Detects fewer faces, but ensures you only blur actual people. For protest footage where anonymity is critical, lower is better. We'd rather blur a mailbox than miss a face. Changing the Anonymization Style By default, Deface uses a gaussian blur. You can also use solid black boxes, which are arguably more secure (blur can sometimes be reversed with enough AI processing power). Why This Matters Clearview AI and other surveillance firms are scraping the web constantly. When you post "raw" footage, you are feeding their dataset. You are helping them map social graphs. Sanitize your media. Strip EXIF metadata (we have a tool for that). Blur faces. Remove audio if not necessary. Use a burner account to upload. Stay safe out there. Open Deface Command Generator