Video Anonymization for Protest Footage: A Safety-First Checklist

Before publishing protest footage, ask whether identifiable participants or bystanders need to appear at all. If they do not, make a separate sharing copy with faces covered and check clothing, signs, voices, and location details too.

A wide crowd scene can contain people with very different reasons for avoiding identification. Do not assume they share the camera operator's willingness to be visible.

Preserve the source separately

Keep the original under the storage and access arrangements appropriate to the project. Make edits to a working copy and label the exported version as treated.

If the recording may support a later investigation or verification process, preserve its original context and consult the responsible organization about handling. This guide is about a public sharing copy, not altering an original record.

Decide whether the scene needs publishing

An image of signs, a speech, or the size of a gathering may communicate the event without close views of participants. Choose the least identifying shot that still supports the story.

Consider people at the edges as well as the central action: passersby, medics, legal observers, and journalists may appear incidentally.

EFF's protest guidance discusses identity and metadata concerns alongside broader safety planning. Adapt any publication decision to the actual circumstances.

Process faces on the appropriate device

If uploading is unsuitable, prepare Unseen's local engine before selecting the sensitive file. You can disconnect after readiness and run the local workflow.

Choose Blur all when no face should remain visible. For local crowded or fast-moving footage, test Maximum and inspect the result. It is not a guarantee that every small or obstructed face will be found.

Use a solid treatment if a softer effect leaves more facial detail than you are comfortable publishing.

Inspect more than faces

Review the image in zones: foreground, background, and both edges. Then inspect reflections and every camera cut.

Look for distinctive clothing, tattoos, names on signs, and people linked together by the sequence of shots. Listen for names or direct conversations. Unseen does not currently remove bodies, arbitrary objects, or audio identifiers.

Use another approved editor for those changes or omit the relevant footage.

Check the final file and its context

Review metadata with an appropriate tool rather than assuming face processing strips it. A title, location tag, or exact timestamp may also disclose information beyond the picture.

Ask another informed reviewer whether the public copy still exposes someone unnecessarily. If you cannot confidently review a dense scene, delay it, shorten it, or choose another shot.

The offline protest review guide explains how to organize the local processing and frame inspection.