Face Blur Online — Free, Fast, Private

How Face Blur Works

This tool does not use a face-detection model — it applies a manual, fixed-position blur. You drag four sliders to define a rectangle (left, top, width, height as percentages of the frame), and that rectangle stays in the exact same place for the full duration of the clip. Under the hood, the tool draws each video frame onto an HTML canvas, clips a second drawing pass to just the rectangle you defined, applies a CSS-style Gaussian blur filter to that clipped region, and layers it back over the original frame. The canvas output is captured in real time with canvas.captureStream() and recorded via the MediaRecorder API into a downloadable video file — the same browser primitives used for screen recording.

Because there's no detection model tracking a face or object across frames, this approach trades automation for predictability and privacy: everything happens locally, with no model weights to download and no video frames sent to a server for analysis. The tradeoff is that if your subject moves around the frame, a fixed box will only cover them part of the time. For a stationary subject — someone sitting at a desk, a static license plate, a whiteboard in a fixed shot — a fixed box works well. For a subject walking through a scene, you'll get better results by trimming the clip into shorter segments and repositioning the box for each one, or accepting a wider box that covers the whole range of motion.

People use this for redacting bystanders in event footage before posting online, hiding a whiteboard or monitor with sensitive information in a screen recording, obscuring a license plate in a dashcam clip before sharing it, or blurring a background detail in a tutorial video. It sits in the /video-tools/ category alongside Add Watermark, which also overlays something onto every frame, except a watermark is meant to be seen and a blur is meant to prevent something from being recognized.

Blur strength matters more than it might seem. A light blur can still leave enough detail for someone to recognize a face if they already know who they're looking for — think of it as obscuring rather than erasing. A stronger blur setting pushes further toward genuine unrecognizability, though it also looks more like an obvious censoring block. If the goal is real anonymization — not just softening a distracting detail — err toward a higher strength setting and review the final export at full resolution rather than trusting the smaller in-page preview.