How do I blur a face in a video?

Upload the video to Valmera and describe the face and the span: "blur the face of the woman on the left for the whole video" or "pixelate the man in the background from 1:10 to 1:35". The AI agent finds the face in the frames, places a censor over it, renders a preview and checks it. You confirm the covered moments and export.

Does Valmera detect and track faces automatically?

No. Valmera has no automatic face detection and tracking. The agent locates the face by looking at frames and places a fixed rectangle for each time window. A seated interviewee takes one request; for someone who moves, it uses a larger box or several windows, and you check the preview before you export.

How do I blur a license plate in a video?

Ask, for example, "black out the license plate on the silver car from 0:12 to 0:20". A parked car in a steady shot takes one rectangle. A car driving through the frame needs a box that covers its path, a few short windows, or a cut.

Is blur, pixelate or a black bar best for anonymity?

A black bar: nothing of the face survives inside it. Pixelation with a generous box is the next safest. Light blur keeps the shape and tone of a face, and blurred faces can sometimes still be recognized, so keep blur for background faces where identity is not sensitive.

Is blurring a face enough for GDPR?

Not on its own. Under GDPR, data is anonymous only when a person can no longer be identified, and a voice, a name badge, clothing or a location can identify someone whose face is hidden. Review the whole video, including the audio, against your requirements, and take legal advice for compliance work.

What about names spoken in the audio?

A censor only changes the picture. Search the transcript to find where the name is said, then ask the agent to cut that passage or mute it. If you added captions, fix or hide the name there too.

Does the blur stay on when I make a vertical version?

Yes. Censors are applied to the original frame before any reframe, so the face stays covered when you crop to 9:16, 1:1 or 4:5. Clips and images you add afterwards are not censored.

Creating an account and uploading are free. Editing requires a subscription. Exports are full-resolution H.264 MP4s rendered from your original file and end with a brief Valmera end card.

Blur Faces and License Plates in a Video

To blur a face or a license plate, upload the video to Valmera and say whose face or which plate, and when: “pixelate the man on the left from 0:10 to 0:45.” The AI agent finds it in the frames, covers it with a blur, a pixelated mosaic or a black bar, then renders a preview and checks the coverage before it replies.

Each censor is a fixed rectangle placed for a time window. Valmera does not detect and track moving faces automatically, so a seated interviewee or a parked car takes one request, while someone walking through the shot needs a larger box, several windows or a cut. You check the preview before you export.

Censor a Face or a Plate

Say what to hide, check the preview, export. Creating an account and uploading are free; editing requires a subscription.

Create an account →

See pricing →

How to Blur a Face or License Plate in Valmera

1
Upload the footage
MP4, MOV, M4V, MKV or WebM, up to 100 GB and 3 hours. Indexing gives the agent a speaker-labeled transcript and frames to look at.
2
Say who or what, and when
“Blur the woman in the red jacket for the whole video.” “Black out the plate on the white van from 0:12 to 0:20.” Mention it if the person moves, so the agent plans several windows.
3
Let the agent place and check it
It looks at real frames, places a rectangle over the face or plate, renders a preview and inspects the covered moments. If anything still shows, it places a bigger box.
4
Check everything, then export
Scrub every moment the person or plate appears, including the edges of each window, and listen for names. Then export the MP4 from Studio.

PROMPTS TO PASTE

Blur the face of the man on the left for the whole interview.

Black out the license plate on the parked car from 0:12 to 0:20.

The cyclist crosses from left to right between 0:31 and 0:36. Pixelate his face in several short windows.

Pixelate her name badge, and cut the two seconds where she says her surname.

Static Interviews vs Moving Subjects

Because a censor holds its place in the frame, how much the subject moves decides the approach.

Situation	What works
Seated interviewee, camera on a tripod	One censor for the whole video, with a box a little larger than the head so a lean or a nod stays covered.
Parked car, or a plate in a steady shot	One censor for the span the plate is visible. Black out or pixelate.
Someone who shifts in their seat or gestures	One larger box that covers their full range of movement.
A person walking across the frame, or a car driving past	Several short time windows, each with its own box, stepping along the path. Or cut those seconds.
Handheld or moving camera	Short windows with generous boxes, checked one by one. If the subject drifts a lot, cut or reframe the shot instead.
Many faces moving at once: a crowd, a busy street	A dedicated face-tracking redaction tool (see below).

A face that is on screen only for a moment is often better cut than censored. For an interview with an anonymous guest, keep the camera locked off when you shoot; a static frame makes one generous box enough for the whole video.

Blur, Pixelate or Black Bar for Real Anonymization

If nobody should be able to recognize the person, use a black bar and make the box generous. Light blur keeps the outline, skin tone and hair of a face, and researchers have trained models that recognized people in blurred and pixelated images (McPherson, Shokri and Shmatikov, Defeating Image Obfuscation with Deep Learning, 2016).

Mode	What survives inside the box	Use it for
Black bar	Nothing. Every pixel inside the box is replaced.	Anonymity that matters: sources, minors, bystanders who did not consent, plates in sensitive footage.
Pixelate	Coarse blocks, about eight across the box’s short side. A bigger box means fewer, larger blocks over the face.	Visible, recognizable-as-censored anonymization in documentary and news-style footage.
Blur	A soft version of the original. Strength scales with the box up to a cap, so shape and skin tone can survive.	Background faces where identity is not sensitive, and cosmetic hiding.

Hiding a face is not the same as anonymizing a person. Under GDPR Recital 26, data is anonymous only when someone can no longer be identified by means reasonably likely to be used, and a voice, a name badge or a location can identify a person whose face is covered. For legal or compliance redaction, review the whole video against your requirements and take qualified advice.

Privacy Checklist Before You Export

Every appearance. The same person can turn up in a wide shot, in the background or after a cut. Check each shot, not just the first.
Reflections. Mirrors, windows, glasses, glossy screens and car paint can show a face the box missed.
Names on people. Badges, lanyards, uniforms, name tags, and labels on mail or packages.
Screens and documents. Monitors, phones, whiteboards and papers on a desk. A black bar is the right mode for text.
Location clues. Plates, house numbers, street signs and shopfronts.
The audio. A spoken name or address. Find it in the transcript, then cut the passage or mute it.
Captions. Burned-in captions repeat what was said, so fix or hide the name there as well.
Added media. B-roll, clips and images you add are not censored.
The exported file. Watch it end to end before you share it.

When to Use a Face-Tracking Redaction Tool Instead

Dedicated redaction software, and some professional editors, detect faces automatically and track them frame by frame. Choose one of those for crowds and street footage with many moving faces, long handheld or body-worn camera shots, and legal or evidential redaction where every frame must be covered and the process documented.

Valmera fits footage where the things to hide are few and mostly still: interviews, talking heads, screen recordings, property tours and vlogs. It also does the rest of the edit in the same chat, so “black out the plate, cut the part where he says the address, and make a 9:16 version” is one request.

Limits

There is no automatic face detection or tracking: the agent places fixed rectangles per time window, and you verify them in the preview. Censors are rectangles only, with no ovals or hand-drawn masks. Timed censors follow the output timeline, so make your cuts first and censor last. Clips and images you add are not censored, and the censor covers the picture only: names in the audio or captions need their own fix.

Related Workflows

To hide screens, documents or anything else in a rectangle, see blur part of a video. To make an object disappear rather than show a censor, use remove an object from a video. Property tours often need house numbers, plates and family photos hidden; see real estate listing videos. For anonymous guests and two-person shoots, see interview editing. Every censor option is in the effects docs.

Creating an account and uploading are free. Editing requires a subscription — see plans.

Using Claude, ChatGPT or Cursor? The same blur_region tool is available through Valmera's MCP server.