How to sync and edit a multi-camera interview
Upload every camera angle and the audio to Vizard Agent and say what the recording is. Vizard Agent lines the angles up, cuts to whoever is talking, and tightens the dead air out of the whole episode. You do not sync anything, set no in-points, and cut nothing by hand.
What is the short version?
Three moves, and the first one used to be an afternoon. Multicam is the format people record and never publish: syncing three files and then cutting an hour of conversation angle by angle is real work, and it has to happen before anyone can even watch the thing.
- Go to Vizard Agent and upload every angle plus the audio.
- Say what it is and how many people are in it.
- Watch it back and correct any cut that landed on the wrong face.
The rest of this page is what goes into each of those, and where the approach stops working.
What do you need before you start?
Every file from the shoot, unedited. The one thing that decides whether this works is whether the cameras share an audio reference — that is what the sync is built from, and it is fixed at record time, not afterwards.
- Audio on every camera. Even a bad scratch track is enough to sync against. A camera recording silence is a camera that cannot be lined up.
- Overlapping recordings. Angles that started and stopped at different times are fine; angles that never overlap cannot be synced to each other.
- The good audio, separately. If you recorded to a field recorder or per-speaker mics, upload that too and say it is the master.
- How many speakers. Two people cut differently from four. Say the number.
What do you type into Vizard Agent?
Vizard Agent ships this as a template, and its prompt asks for exactly the three things that change the edit: how many angles, what kind of recording, and that you want the whole thing tightened rather than just assembled.
Prompt
Three variants worth keeping:
When one file is the good audio
When you want a wide shot to fall back on
When you want clips as well
What are the steps inside Vizard Agent?
Three, and the middle one carries the information. Vizard Agent aligns the angles from their audio, transcribes the conversation, decides who is speaking when, cuts accordingly and removes the dead air — all in one request, so what comes back is a watchable episode rather than a synced timeline.
- Upload every file to the Vizard Agent chat with the prompt above.
- Pick the tier that matches the job. For anything published under a show name the product recommends Max; the free Flash tier is built for quick, simple edits.
- Watch it at full speed. Sync errors are obvious in motion and invisible when scrubbing; the reverse is true for cut choices.
When does this not work well?
Multicam depends on two things: a shared audio reference to line the angles up, and clean enough speech to tell who is talking. Everything that fails here fails at one of those, and both are decided while recording.
- A camera with no audio. Nothing to sync against. If a camera was set to record silent, it has to be lined up by hand elsewhere.
- Angles that never overlap. A camera that started after the other stopped cannot be synced to it — they are two recordings, not two angles.
- Heavy crosstalk. Two people talking over each other on one shared microphone can be hard to attribute, and a cut to the wrong face is more distracting than no cut at all.
- More than three or four speakers. Every extra person multiplies the who-is-talking decisions, and a round-table changes speaker faster than a cut can comfortably follow.
- Cameras with very different looks. Angles shot on different bodies, white balances or exposures still cut together, but the mismatch will be visible unless you ask for a matching pass.
- Drifting clocks on long records. A two-hour record from cameras with different clocks can drift; if the back half looks out of sync, say so and give the timecode.
If one angle is unusable, say which and ask for a single-camera edit instead — a bad cut to a bad angle is worse than staying wide.
What does the result look like?
One assembled episode: angles synced, cuts on whoever is speaking, dead air removed. Vizard Agent keeps every source file and the transcript in the project, so pulling short clips from the finished episode later costs a sentence instead of another upload of several gigabytes.
It is not instant, and this category was not separately measured. The nearest measured work — finding and cutting moments inside long recordings — runs a median of about 38 minutes end to end, with the middle half between 18 and 61 minutes; a multi-angle upload sits at the upper end because there is simply more footage to get through. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.
How do you fix a result that came back wrong?
Reply in the same Vizard Agent conversation instead of starting over. Every angle and the transcript are still in the project, so you never re-upload several gigabytes of footage and you are not charged again for the sync and the transcription Vizard Agent has already done.
| What you see | What to say next |
|---|---|
| It cuts to the wrong person | "At 04:12 that is me talking, not her — stay on camera B" |
| Too many cuts | "Cut less — hold each angle for at least five seconds" |
| Sync drifts later on | "The angles drift apart after about 40 minutes, re-sync the back half" |
| One angle looks wrong | "Match camera B's colour to camera A" |
| Too tight | "Leave the natural pauses between questions" |
How does Vizard Agent compare to editing it yourself?
By hand, syncing comes before anything else: line up the waveforms, build a multicam clip, then watch the whole episode making a cut every time somebody speaks. The syncing is mechanical and the cutting runs at real time, which is why a one-hour episode costs the better part of a day.
| Vizard Agent | Editing multicam by hand | |
|---|---|---|
| Syncing angles | Automatic, from the audio | Waveform align, then verify |
| Cutting to the speaker | Automatic | A cut at a time, at real time |
| Removing dead air | Part of the same request | Another pass |
| Time for a 1-hour episode | Tens of minutes | Most of a day |
| Control over each cut | Directed by language | Frame-level |
Neither replaces the other. Vizard Agent wins on the mechanical work and on keeping a weekly show weekly; an editor still wins when the rhythm of the cutting is the craft.
Common questions
Do the cameras need timecode sync?
No. Vizard Agent lines the angles up from their audio, which is why every camera needs to have recorded some — even a poor scratch track from the on-board mic is enough. Proper timecode helps but is not required.
Can I use a separate audio recorder?
Yes, and you should if you have one. Upload it and tell Vizard Agent it is the master audio: the cameras sync to it and the episode uses the good recording rather than a camera mic.
What if one camera stopped mid-episode?
Say so. Vizard Agent works with what overlaps and stays on the remaining angle for the rest, which is what a human editor would do — just tell Vizard Agent rather than letting it discover the gap.
Can I get short clips from the same upload?
Ask Vizard Agent in the same conversation. The footage and the transcript are already in the project, so clips cost a sentence rather than another upload — there is a separate guide on clipping linked below.
Does it fix mismatched colour between cameras?
Ask Vizard Agent for it explicitly — "match camera B to camera A". Without the instruction the angles are cut as shot, which is fine for similar bodies and visible when the cameras were not the same.