Surveillance Guide

How to Redact a Video Blur Faces, License Plates, and Audio for FOIA and Disclosure

Redacting a video means permanently obscuring everything in the footage that identifies a person or a private detail: faces, license plates, names, badges, screens, and spoken identifiers in the audio. The reliable way to do it is to detect those items, mask them so the cover follows each subject across every frame, redact the audio, and export a new reviewed copy while keeping the original evidence untouched. Doing it by hand works but takes hours per clip, which is why most teams now let AI handle the detection and a person handle the sign-off.

Start Free Trial View Pricing

The Short Answer

What Redacting a Video Actually Involves

To redact a video you cover up the parts that reveal a person's identity or a protected detail so the released copy can be shared without exposing them. In practice that is faces, license plates, house numbers, name tags, computer and phone screens, documents held up to camera, and the voices or spoken names in the audio. You apply a blur, pixelation, or solid block over each item, and that mask has to track the subject as it moves, frame by frame, or the person flashes back into view the moment they turn or the camera pans.

The other half of the job is doing it defensibly. A proper redaction never alters the source recording. It produces a separate, flattened export where the masks are burned in and cannot be peeled back off, and it logs what was changed and by whom so the chain of custody holds up if the release is challenged. That audit trail is what separates a courtroom-ready redaction from a quick blur that an opposing attorney can pick apart.

You can redact short clips manually in a video editor, but it is slow and error prone. AI redaction tools detect the faces, plates, and other identifiers automatically and follow them across the footage, cutting the manual effort by more than 90 percent. A trained reviewer still confirms the result, which is the workflow this guide walks through.

At a Glance

Redact in videoFaces, plates, screens

Redact in audioNames, identifiers

Mask mustTrack every frame

Original fileStays untouched

Manual timeHours per clip

AI time saved~90%+

General guidance, not legal advice. Follow your agency or counsel's release rules.

Know the Targets

What You Need to Redact in a Video

Redaction is more than blurring the obvious face. Bystanders, reflections, and on-screen text all carry identifying information. Miss one and the release can expose someone you were obligated to protect.

What to redact	Why it matters	Typical treatment
Faces of bystanders and witnesses	Uninvolved people have a privacy interest in not being identified	Blur or pixelate, tracked per frame
License plates	A plate links footage to a registered owner and home	Blur, pixelate, or solid fill
Names, badges, tattoos	Name tags and distinctive marks are personally identifying	Blur the region
Screens and documents	Monitors, phones, and paperwork expose data and case detail	Solid fill or heavy blur
House numbers and locations	Addresses can reveal where a victim or minor lives	Blur the region
Audio: names and identifiers	A spoken name or address identifies someone as clearly as their face	Mute or bleep the segment

What you must redact depends on the law that governs the release (public records statutes, CJIS, HIPAA, FERPA, GDPR) and your agency's policy. When in doubt, redact and document the exemption you relied on.

The Workflow

How to Redact a Video, Step by Step

This is the repeatable workflow whether you redact a body-cam clip, a CCTV export, or a dash-cam file. AI does the detection; you keep control of the review and the export.

01

Work from a copy, never the original

Start by importing the footage into your redaction tool, which works on a copy and leaves the source file intact. The untouched original stays in evidence so the chain of custody is never broken, and you can always re-redact if a different exemption applies later.

02

Auto-detect faces, plates, and PII

Run automatic detection. The software scans the video and flags every face, license plate, and identifier, then attaches a mask that follows each one as it moves through the frames. On a busy scene this finds people a manual pass would miss.

03

Review the timeline and fix misses

Scrub the timeline and check the detections. Add a mask anywhere the AI missed an item in low light, a crowd, or an odd angle, remove any false catch, and confirm each mask stays locked on its subject from first appearance to last.

04

Choose a redaction treatment

Pick how the masked area looks: blur, pixelate, or a solid fill. Use one style consistently across the clip unless the release instructions say otherwise, because a mixed output makes the quality review harder and can look inconsistent on release.

05

Redact the audio track

Video is only half the recording. Mute or bleep the segments where a name, address, phone number, or other identifier is spoken. Skipping the audio is one of the most common ways an otherwise clean visual redaction still leaks a protected identity.

06

Export a flattened copy and log it

Export a new file with the masks burned in so they cannot be reversed, keep the original untouched, and save the audit trail of what was redacted and by whom. That record is your defense if the release is questioned or appealed.

Automatic vs Manual

Should You Redact by Hand or With AI?

Both methods produce a valid redaction. The difference is time, consistency, and how well it holds up at scale. For most teams the answer is a hybrid: let AI do the heavy lifting, then have a person sign off.

Manual redaction

In a standard video editor you pause, draw a blur or block over each item, then nudge it frame by frame so it tracks the subject, much like rotoscoping in film. It gives precise control and works fine for a single short clip. The problem is time: one recording can take hours, the effort scales with every release, and a tired reviewer eventually misses a face in the background. It does not hold up well against a growing backlog.

AI redaction

AI detects faces, plates, and identifiers automatically and applies motion-aware masks that follow each subject across frames, cutting manual effort by more than 90 percent and staying consistent across long footage. It still needs a human pass, because edge cases like low light, heavy crowds, obstructions, and non-standard identifiers can fool detection, and a courtroom may expect confirmation that an officer reviewed the result. Treat the AI as the first pass, not the final word.

Keep the Original, Burn In the Copy

Whichever method you use, the rule that protects you is the same: never modify the source recording. A redaction workflow that edits the original, or that cannot show what was changed and by whom, creates a chain-of-custody risk that can sink the whole release. The original stays sealed in evidence, the export is irreversible, and the tool generates an audit trail and compliance record automatically.

One more legal note specific to public records. Blurring a license plate or a face is not automatically a FOIA violation, as long as the agency is applying a valid privacy exemption and can justify it. What gets agencies in trouble is blanket redaction with no stated reason, which invites appeals and costly challenges. Redact what the law lets you withhold, document the exemption, and release the rest.

Scale It Up

When Redaction Becomes a Volume Problem

A one-off clip you can redact in an afternoon. The trouble starts when redaction becomes routine: every public records request, every body-cam release, every camera export that has to go to a defense attorney or an insurer. That is when manual editing stops scaling and a dedicated tool earns its place. Purpose-built video redaction software turns a multi-hour job into a guided review, with the same AI detection and audit trail on every file.

The detection underneath redaction is the same technology that powers the rest of a modern video platform. Facial detection finds the faces to mask, and license plate recognition locates the plates. When you need to pull the right clip before you can redact it, video forensics tools help you find and prepare the footage. And because so many redactions exist to satisfy a disclosure rule, the same controls support government and public-sector releases and broader privacy compliance.

If you are redacting to honor a retention or disclosure deadline, it helps to know how long the underlying footage should exist in the first place. Our guide on how long to keep security camera footage covers the retention windows that drive most release schedules.

FAQ

Common Questions About Redacting a Video

How do you redact a video?

You redact a video by importing a copy into a redaction tool, detecting every face, license plate, and identifier, and applying a blur, pixelation, or solid block that follows each item across the frames. Then you redact the audio for spoken names and details, export a new flattened copy with the masks burned in, and keep the original untouched. AI handles the detection and a person reviews the result.

How do you blur faces in a video?

To blur faces in a video, use a tool that detects each face and attaches a blur or pixelation mask that tracks the face as it moves and turns. Automatic detection finds faces frame by frame, including bystanders in the background, while a reviewer adds masks for any the software missed in low light or crowds. The blur is then burned into an exported copy so it cannot be removed.

How do you redact a license plate in a video?

Redact a license plate by running plate detection over the footage, confirming the tool catches the plate every time the vehicle is visible, and applying a blur, pixelation, or solid fill that tracks it across frames. Check the timeline for moments the plate reappears after being off screen, fix any misses manually, then export a reviewed copy. The original recording stays intact in evidence.

How long does it take to redact a video?

By hand, a single video can take hours because you adjust the mask frame by frame. AI redaction cuts that by more than 90 percent: detection runs automatically and a person spends their time reviewing rather than drawing masks. Exact time still depends on the length of the clip, how many people and plates appear, and how complex the scene is, but a job that took an afternoon often drops to minutes of review.

Is there free video redaction software?

You can blur regions in free video editors, but they have no automatic detection, so you draw and track every mask by hand, and they rarely produce an audit trail or guarantee the original stays unaltered. For one short clip that may be enough. For routine releases, FOIA backlogs, or anything that has to hold up legally, dedicated redaction software pays for itself in saved time and defensibility.

Is blurring a license plate a FOIA violation?

Not automatically. Blurring a license plate or a face is allowed when the agency applies a valid privacy exemption and can justify withholding that detail. The problem comes from blanket redaction with no stated reason, which can trigger appeals and legal challenges. Redact what the law permits, record which exemption you relied on, and release the rest of the footage.

Can you redact audio in a video?

Yes, and you usually have to. A spoken name, address, or phone number identifies a person as clearly as their face, so a complete redaction mutes or bleeps those segments of the audio track. Tools built for body-cam and CCTV redaction handle audio alongside the video. Forgetting the audio is one of the most common reasons a visually clean redaction still leaks a protected identity.

Explore More

Stop Redacting Footage Frame by Frame

Surveillant connects to the cameras and footage you already have and adds AI detection of faces, plates, and identifiers, the same technology that powers fast, defensible video redaction. Start a free 14-day trial and see how much time automatic detection saves.

Start Free Trial Explore Video Redaction

General guidance only, not legal advice. Follow your agency and counsel's release rules.