Home / Video redaction software

Video Redaction Software

Detect faces, licence plates, screens, weapons and anything else you draw, tracked across the whole clip. Review every detection, then burn the redaction into a new rendition. Runs in your browser, or entirely inside your own network.

FACE FACE PLATE IN-CAR SCREEN FRAME 0412 / 1800 SAME OBJECT, TRACKED

Why Agencies Move

62 days to 7

Average video release time at a county public safety agency

1.1M

Recordings redacted in a single deployment

315

File formats in, no conversion step

What Makes Video Hard

30 frames every second

A face is not redacted once, it is redacted 1,800 times a minute. Miss one frame and the release is a disclosure.

Objects do not cooperate

A face turns away and comes back as what the detector reasonably believes is a different person. Someone walks behind a pillar. A plate is legible for eleven frames out of four hundred.

One scene, several classes

A body-cam frame holds faces, a plate, the terminal in the patrol vehicle, a house number and a weapon, each needing different handling in the same shot.

360 footage breaks flat tools

A tool that treats an equirectangular frame as a rectangle puts the box in the wrong place. Detection converts between the projection and spherical coordinates so the mask lands where the object actually is.

Desktop tools do not scale

One analyst, one machine, one file. Nothing shared, nothing tracked, and no supervisor able to see what was redacted or by whom. Redactor runs centrally in the browser.

Formats nobody planned for

Footage arrives from whatever produced it, including decommissioned camera systems and twenty-year-old archives. 315 extensions are handled without a conversion step.

Metadata survives redaction

Video files carry GPS coordinates, device identifiers and timestamps. A release that masks a face and ships the coordinates has not protected anyone. Metadata is stripped or rewritten on export.

What You Can Redact

Faces and people

Persons, faces, heads and body parts

Licence plates

Detected and read as text, so a plate is searchable

Vehicles

Cars, buses, trucks, bikes, boats, aircraft and trains

Weapons

Firearms and other weapons

Screens and devices

Monitors, laptops, phones and in-car terminals

Identity documents

ID cards, driving licences, passports and name tags

Tattoos

Identifying body art that survives a face redaction

Street signs

Signs and house numbers that identify a location

Spoken PII

Names, addresses and numbers on the audio track

How It Works

01

Analyse

Detection runs at upload or on demand. Scope is selectable by duration and by class, so part of a video processes on its own.

02

Review

The studio shows each object as a track across the timeline. Merge duplicates, split, adjust, delete false positives, draw what was missed. On MP4 sources tracking runs in the browser, so a drawn object is followed immediately with no server round trip.

03

Style

Blur, pixelate or solid fill, chosen per job. Greyscale export where colour itself is identifying.

04

Produce

The redaction is burned into a new rendition. The original is governed by a configured policy, not a side effect.

Redacting Some of It, Not All of It

Keep one subject visible

Every other face is masked while the person under review stays in frame, which is what use-of-force review and internal affairs actually need.

Redact part of a timeline

An object is obscured for the stretch of its time on screen that matters, rather than for the whole recording.

Serve several audiences from one asset

Redactions organise into named layers, so a court copy and a public release come from the same file instead of separate copies that drift apart.

Proof

Average video release time

62 days → 7 days

Missed-deadline notices

~50 a month → ~12 a month

Investigator effort per case

~8 hours → 45 minutes

Prosecutor packages

3 business days → same day

A county public safety agency with more than 200 sworn officers and 175 body-worn and in-car cameras was taking about 150 public information requests a month, with two staff handling disclosure for all of it. Video and audio were 28% of the requests and 76% of the redaction hours, because a typical request runs five to ten hours of work per hour of footage and a crowded scene runs fifteen to thirty. One request, a 30-minute foot pursuit on a single camera, took eight to ten hours because the footage caught licence plates, house numbers and bystanders throughout.

Redactor was deployed on the agency's own infrastructure, alongside the cloud evidence management system it already ran. Nothing was replaced. The agency has met every date it has set since.

Read the full case study

Compared With a Video Editor

Frame-by-frame editing Redactor
Draw a box, keyframe it, repeat for every object Detect once, correct the track, done
Audio handled in a separate tool Spoken PII in the same pass
One analyst, one desktop, nothing shared Central, browser-based, supervisor-visible
No record of what was removed or why Exemption code on every redaction, printed on the output
A mask layered over the original Burned into a new rendition

Where It Runs

Shared SaaS, dedicated cloud, your own private cloud, on-premises, hybrid, or fully air-gapped with the entire AI stack local and no outbound inference call.

Deployment options

FAQ

Video Redaction Software questions, answered

Can redacted video be un-redacted?

No. The redaction is burned out of a new rendition rather than layered over the original, so the obscured content is not present in the released file and cannot be recovered from it. What happens to the source file is a separate, configured decision.

How long does it take to redact an hour of footage?

Manual frame-by-frame work typically runs five to ten hours per hour of footage, and fifteen to thirty for a crowded scene. With detection doing the first pass, the operator time becomes review time rather than drawing time. One agency measured investigator effort per case falling from about eight hours to 45 minutes.

Can I keep one person visible and redact everyone else?

Yes. Selective redaction keeps a designated person unmasked while every other face in the scene is obscured, tracked through movement and occlusion. It is the standard requirement for use-of-force review and internal affairs.

Does it redact the audio as well as the picture?

Yes. Spoken PII on the audio track is detected and muted or bleeped in the same pass, so a request does not need a second tool for the sound.

Will it work on footage from an old camera system?

315 file extensions are handled on ingest without a conversion step, including proprietary CCTV output and legacy archives from decommissioned systems.

Can it run without sending footage to the cloud?

Yes. Redactor runs on-premises or fully air-gapped, with the entire AI stack local — no content and no inference request leaves the deployment.

See It on Your Own Footage

Send us something difficult — a crowd, poor light, a screen in frame, a 360 camera. That is the useful test, not a clean clip.