How to anonymise one person in a video and keep the other named
Tell Vizard Agent who is protected and from what, person by person. Anonymity is rarely all-or-nothing: the practitioner usually wants her name and title on the clip, the clients agreed to appear but not to be identified, and that means replacing an on-screen name while leaving faces exactly as they are.
What is the short version?
You have a recording of a private session and you want to post highlights. One person on screen is the professional and wants credit; the others are clients. Vizard Agent protects each of them to the extent they agreed to.
- Tell Vizard Agent who may be identified and who may not.
- Say what they consented to — faces, voices, names.
- Have every shot with a name checked, not just the first.
What do you need before you start?
The clip, and a clear account of what was agreed. "They are happy to be seen but not named" is a complete specification and it produces a very different edit from "anonymise them" — the second would mask the faces, and the faces are the reason the clip works.
Know what the material is. A recording where people are emotional is not a product demo, and the sensible default is the lightest possible intervention: change what identifies them and touch nothing else.
What do you type into Vizard Agent?
Name each person and say how each is treated. It reads like over-explaining and it is in fact the whole instruction — a single word like "anonymise" cannot express that one person's name stays on screen while another's comes off, which is exactly what Vizard Agent needs to know.
Keep the reading as it is, only trim the gaps. Remove the family's names but keep my name and title. They have consented to their faces being shown.
That last sentence is doing real work. Without it, Vizard Agent has to assume the stricter reading and mask the faces, which would remove the reactions that carry the clip.
What does Vizard Agent actually do?
It finds where each person is identified and treats each case separately. In this kind of footage the identification is usually an on-screen label rather than anything spoken, so the fix is a replacement rather than a blur — the family's name label became "Parents", and the practitioner's own identification stayed and later gained her full professional title.
In that session the work ran:
- Confirmed what was consented to before touching anything, and left the faces visible on that basis.
- Located the family's name label and inspected it at full size.
- Replaced it with a neutral word rather than covering it with a box.
- Checked every shot of the parents so no name remained visible between cuts.
- Inspected a framing change near the end, where the label moves, and confirmed its exact position in the final shot.
- Checked the edge of the original label, since a replacement that is a pixel short leaves a sliver.
Then the edit itself, which was deliberately almost nothing: a two-second trim on the longest pause and no change at all to the spoken reading.
What does the result look like?
The same clip, with the clients unnamed and the practitioner properly credited. Their faces and their reactions are intact, the dialogue is untouched from end to end, and the only structural change Vizard Agent made was shortening one long pause.
The captions get verified against the audio rather than trusted. In that session a repeated phrase was checked twice against the original recording to confirm it was what the speaker actually said and not a caption error — on material this sensitive, putting a word in someone's mouth is the serious failure, not a typo.
When does this not work well?
When the name is spoken rather than shown. Cutting a name out of continuous speech leaves a hole, and muting it is audible; the options are a cutaway over it, a short bleep, or accepting it. Vizard Agent will say which is least damaging to the sentence.
It also cannot help when the person is identifiable by context. A story that names a town, a date and a relationship identifies a family to anyone who knows them, and no amount of label replacement changes that. That is a judgement about publishing, not an edit.
And consent can be narrower than you remember. Agreement to be recorded is not agreement to be posted, and it is worth confirming before the clip goes out rather than after.
How do you fix a result that came back wrong?
If a name survived somewhere, give the timestamp. Labels move when the framing changes, so a replacement that is correct for most of the clip can miss the one shot where the camera repositioned — Vizard Agent re-checks each shot rather than the whole run.
If the replacement looks like a patch, say so. A neutral word set in the original's own style reads as a design choice; a grey rectangle reads as a redaction, which draws the eye to exactly what you were hiding.
If you decide the faces should be masked after all, say that as a change of scope. Vizard Agent re-does the protection rather than adding a blur on top of a cut it built on the assumption faces were fine.
How does Vizard Agent compare to doing it yourself?
The hard part is not the replacement, it is the coverage. You will fix the label in the shot you were looking at, export, and find it again later in the clip where the camera moved — and on sensitive material that mistake is published before you notice.
Checking every shot the label appears in, and the exact edges of it, is mechanical work that Vizard Agent does as a pass rather than as a spot check. The other thing it adds is asking what was consented to before deciding how much to protect, which is the question that determines whether the clip is usable at all.
Common questions
Should I blur the faces too? Only if that is what was agreed. Vizard Agent protects to the scope you state.
Why replace the name rather than cover it? A neutral word looks intentional. Vizard Agent matches the original label's style.
What if a name is spoken aloud? Vizard Agent will propose a cutaway, a bleep or leaving it, and say what each costs.
Can the practitioner keep her title? Yes. Vizard Agent keeps her identification and can add the full title.
Will the dialogue be edited? Only if you ask Vizard Agent to. Here it was untouched apart from one pause.
How does it find every instance? By checking each shot rather than sampling. Labels move when framing changes.
Can it make a vertical version? Yes, with the protection carried over. Ask Vizard Agent in the same job.
Are the captions accurate? Vizard Agent checks doubtful lines against the original audio rather than trusting them.
What about a voice being recognisable? Say if that matters. Altering a voice is possible and changes the performance.
Can it remove a name from a document on screen? Yes, the same way, and it verifies every frame the document is visible in.
What if I only want part of the clip public? Say which part. Vizard Agent cuts to it and applies the protection there.
Does the original file change? No. Vizard Agent delivers a new file and leaves the source alone.
Can I get a version with no identification at all? Yes, as a separate deliverable with a wider scope.
How long does this add? Minutes. The checking pass is the slow part, and it is the point.