How to point out which person in the video is you
Describe yourself to Vizard Agent by something visible and something positional — "the green belt, on the referee's left" — rather than expecting it to recognise your face. Then say what counts as your moment: in most recordings you are on screen far more often than you are doing anything, and only the second kind is worth cutting.
What is the short version?
Face recognition is the obvious approach, and it needs a clear, front-on, reasonably large face to work from. A competition recording hands Vizard Agent the back of a head, a helmet, or two people in identical kit at twenty metres from the camera.
- Give Vizard Agent a colour and a position, not a name.
- Say which actions count as your moments.
- Check the shortlist before the edit is built.
What do you need before you start?
The recording and a description you would give a friend. Vizard Agent works from exactly the kind of detail people use in real life — what you were wearing and where you were standing relative to something that does not move.
- The recording. The full session or match.
- What you were wearing. Colour first.
- Where you were. Relative to a referee, a net, a corner.
- What counts as a moment. Throws, saves, goals, solos.
- Whether you change sides. Halves and rounds matter.
What do you type into Vizard Agent?
Anchor your description to something that does not move. "On the left" stops meaning anything the moment the camera shifts; "on the referee's left" survives any camera move, because Vizard Agent can see the referee in the frame as well.
Prompt
Variants worth knowing:
- "On the referee's left." A position that survives a camera move.
- "My active moments, not every time I am on screen." The real filter.
- "Show me the shortlist first." Cheap to correct at this stage.
What does Vizard Agent actually do?
Here is the Vizard Agent sequence on a real competition recording cut down to one competitor. Identification and selection run as two separate passes, and collapsing them into one is exactly what produces a highlight reel of somebody standing about.
- Measures the recording and its basic properties.
- Builds a quick visual map of the whole session.
- Reviews the combined frames to see who is where.
- Finds the active moments of the competitor in the green belt.
- Builds a fuller frame map to place those moments in context.
- Pins down the throws specifically.
- Pins down the strongest groundwork sections.
- Prepares the candidate frames for review.
- Shows the selected actions as a preview before cutting.
- Cuts the chosen moments to the music's beats.
Step four is the one that needs your description. A kit colour plus a positional anchor is enough to follow one person through a session in a way a face signature cannot when the faces are turned away.
Steps six and seven are a second filter, and they are what make the reel watchable. "Active" is still too broad on its own; naming the kinds of action you want — throws, groundwork — turns forty appearances into eight moments.
Step nine is worth insisting on. Confirming the shortlist takes a minute and saves rebuilding an edit that followed the wrong person in green.
What does the result look like?
A reel of you doing things. Not you in the background of somebody else's match, not you waiting between rounds — the moments where you were the one acting, and Vizard Agent shows the shortlist so you can say which ones you meant.
When does this not work well?
Some recordings cannot be disambiguated at all. Two competitors in the same colour, a wide shot where everybody is a few pixels tall, or a camera that never once establishes the fixed reference you described will each defeat the method, and Vizard Agent will say so rather than guess.
- Identical kit. Colour stops distinguishing anyone.
- Very wide shots. Nobody is big enough to follow.
- No fixed reference. Nothing to be positioned relative to.
- Sides that swap. Your anchor reverses at half time.
- Long gaps off camera. There may be less of you than you think.
How do you fix a result that came back wrong?
Add a second identifying detail to the first one. A single description usually gets Vizard Agent most of the way there, and the corrections that follow are almost always a matter of ruling out the one other person in the recording it kept confusing you with.
- "That is the other guy." A second detail is added to tell you apart.
- "These are just me standing there." The action filter is tightened.
- "You missed the second round." The anchor changed; sides are re-stated.
- "Too short." The action list is widened to include more kinds of moment.
How does Vizard Agent compare to doing it yourself?
By hand you scrub the whole recording looking for yourself, which works and takes as long as the match did. Automatic face tracking is faster and, in a sport, spends most of the recording confidently following somebody else.
| By hand | Vizard Agent | |
|---|---|---|
| Identifying you | By eye, throughout | From a colour and an anchor |
| When faces are hidden | Still possible, slowly | Still possible; faces not needed |
| What counts as a moment | Decided as you go | Named up front |
| Wrong person | Found at the end | Caught on the shortlist |
| Effort | The length of the recording | A description and a check |
Common questions
Why not just use my face? Because sport rarely shows it clearly. Vizard Agent uses kit and position instead.
What if several people wear the same colour? Add a second detail — a number, a position, a hairstyle.
What is a positional anchor? Something in frame that stays put — a referee, a net, a corner. Vizard Agent measures from it.
What if I change sides? Tell Vizard Agent when, and it switches the anchor at that point.
Can it tell active from standing? Yes, if you name the actions. Otherwise Vizard Agent counts every appearance.
Should I see the shortlist? Yes. It is the cheapest moment to catch a misidentification.
Does this work for non-sport? Yes — a conference, a wedding, a group performance. The method is the same.
What if I am only in wide shots? Vizard Agent will say so. There may be less usable material than you expect.
Can it follow two people? Yes. Describe each one and Vizard Agent tracks them separately.
Will it cut to music? Yes, once the moments are agreed. That is a separate step.
Can I add more moments later? Yes. The shortlist is a list; extending it does not mean starting again.
What if the camera never shows the referee? Then pick another fixed thing. Vizard Agent needs something stationary.
Does the video need to be high quality? It helps, but colour and position survive worse footage than faces do.
Can it prove it followed the right person? Yes — that is what the shortlist frames are for.