Redact faces from video, automatically and irreversibly
Upload a video and every face is detected, tracked across frames, and permanently removed by re-encoding the pixels — not an overlay, not a reversible mask. The result is a new video file where the original face data no longer exists.
Redacting faces from video means permanently destroying the pixel data that makes a person identifiable in every frame of the recording — not applying a removable filter, not overlaying a black box on a thumbnail. True face redaction writes a new video file in which the original facial detail no longer exists.
This page explains how automated face redaction works, why it outperforms manual redaction for legal and compliance contexts, and which real-world workflows depend on it. You can redact faces from a video right now with no account.
How face redaction software works
Reliable video redaction combines two distinct passes:
- AI detection identifies face regions in individual frames using a neural model.
- Deterministic re-encoding replaces those pixel regions with obscured data and writes an entirely new video file.
Between those two steps, a tracking layer interpolates face positions across frames so the redaction stays locked on each subject even during motion blur, partial occlusion, or rapid movement. Without tracking, per-frame detection alone leaves gaps — a briefly turned head or a fast pan can expose a face for several frames, which is unacceptable in legal or compliance contexts.
The critical distinction is what happens to the pixels: Medianonymizer does not apply a visual overlay that sits on top of the original. It re-encodes the face regions directly into the output bitstream. The source pixel values are gone. There is no project file, no mask track, and no way to recover the original frame underneath.
This is the difference between obscuring a face and redacting it.
Why automated redaction beats manual video editing
Manual face redaction in a video editor is impractical at any meaningful scale:
- A two-minute clip at 30 fps is 3,600 frames. Keyframing a redaction box across all of them takes hours of editor time per minute of footage.
- Human reviewers miss background faces, reflections, and faces in fast motion — precisely the frames most likely to leak identity during legal review.
- Most "blur" or "mosaic" effects in non-linear editors are non-destructive overlays. The original frame is still present in the project; the effect is reversible and the source footage remains intact — which fails any audit requiring irreversibility.
Automated face redaction software eliminates all three failure modes: it processes every frame consistently, catches faces that a human would miss, and writes a new file from which the original data is absent.
Real-world use cases for video face redaction
Law enforcement and legal proceedings
Body camera footage, interview recordings, and courtroom videos routinely contain faces of witnesses, minors, or third parties who are not parties to the case. Redacting those faces before disclosure or publication is a procedural requirement in many jurisdictions. An auditable output — where the redaction method is documented and the result is irreversible — satisfies the evidentiary chain of custody better than an editor overlay would.
Corporate investigations and HR
Internal investigation recordings, exit interviews, and surveillance footage may need to be shared with external counsel or regulators while protecting the identities of employees not under review. Face redaction software that produces an auditable, irreversible output reduces legal exposure when handing over evidence.
Research, healthcare, and clinical video
Patient-facing video — consultations, therapy sessions, procedure recordings — cannot be shared for training, auditing, or research without removing faces. Automated redaction makes it feasible to process large collections of clinical video while maintaining a per-file audit trail required under HIPAA or equivalent frameworks.
Redaction is not just about faces
A person can be identified in a video through their face, but also through license plates in the background, visible ID documents, name badges, and spoken personal data in the audio track. A complete redaction workflow addresses all of these in the same pass. Medianonymizer can redact faces, blur plates, and beep or mute sensitive audio segments in a single job — so you produce one output file with one audit record, not several partial redactions stitched together.
Redact faces from your video now
Upload a video, select face redaction (and plates or audio if your workflow requires it), review the exact price, and download an irreversibly redacted file. No account required, no subscription — pay only per job.
Frequently asked questions
- Can the redaction be reversed — is the original face recoverable?
- No. Medianonymizer re-encodes the pixels in each face region before writing the output file. There is no mask layer, no hidden track, and no reference to the source frames underneath. Once the job is complete, the original face data is gone from the output. This is what distinguishes deterministic re-encoding from reversible overlay tools, which can be peeled back.
- Is face redaction in video sufficient for GDPR compliance?
- Permanently redacting faces removes a direct biometric identifier, which is a required step when anonymizing video under GDPR. For the footage to be fully out of scope, you should also address other personal data in the same clip — spoken names, phone numbers, visible ID documents, and license plates. Medianonymizer can redact faces, plates, and sensitive audio segments in a single job, giving you a complete audit trail for each file.
- What video formats does the face redaction support?
- Uploads in MP4, MOV, WebM, and MKV are supported. The redacted output is delivered as a standard MP4 re-encoded with the face regions permanently altered. There is no hard resolution cap for typical footage; longer or higher-resolution videos take more processing time and are billed per minute of duration.
- Can I redact faces from many videos at once?
- You can run multiple jobs in parallel by uploading files one after another — there is no queue and no account required. For high-volume workflows or programmatic access, contact us about API integration. Each job is priced individually so you pay only for what you process.
- How much does video face redaction cost?
- Video redaction starts at €3.00 per job plus €0.05 per minute of footage. The exact price is shown before you confirm payment — no subscription, no account, no hidden fees.