How to Blur Faces in Low-Light Video

Low-light video needs extra attention where the light changes. A person who is unrecognizable in a dark frame can become clear for a moment under a streetlight, phone screen, or camera flash. Blur those appearances rather than assuming the whole shot is too dark to identify anyone.

In Unseen, start with the clearest source file you have and inspect the bright moments first.

Keep the camera original if you have it

A messaging-app copy may have less detail than the camera file. That can make a face harder for the detector to find without making it unrecognizable to someone who knows the subject.

Avoid deliberately darkening the video as a substitute for masking. Viewers can change brightness, and other frames may contain a clearer view.

If you need a color-corrected version for the final edit, prepare that version before the face pass. This lets you assess the appearance you intend to publish.

Test the flashes of visibility

Mark moments where a subject turns toward a lamp, crosses a shop window, or lifts a bright phone. Also inspect the frames immediately before and after a flash. The sudden contrast change may produce inconsistent detections.

Use Select faces if particular people should stay visible. If all people should be hidden, Blur all avoids a separate selection decision. Neither path guarantees that a dark face will be found.

For local processing, Maximum is worth testing when clear appearances are brief. It increases detection frequency; it does not recover face detail that the camera failed to capture.

Choose a treatment you can evaluate

Soft blur can be difficult to distinguish from noise in a dark preview. Pixelate or Black bar makes it easier to see whether a treatment is actually present.

Watch coverage separately from appearance. A black region that follows the head is useful. A strong effect left behind as the person moves is not.

If a face keeps escaping detection, draw a box on it in Review & fix faces and track it (step by step), or cut or crop that interval. Apply the correction across the full interval where the person could become recognizable, not just the single brightest frame.

Review at the intended brightness

Watch the exported file on more than one display if possible. A bright phone screen can reveal details that seemed absent on a dim laptop. Include reflections and distant people illuminated by the same light source.

For a night-out recap, the most revealing moment may be someone calling a name while the camera pans toward them. Face processing does not edit that audio.

Use the missed-frame guide to decide whether to rerun detection or move to a different correction.