How to cut a music video from your performance takes

Last updated 2026-08-25 · 7 min read

What is the short version?

A music video cut from performance takes has one hard constraint that no other edit has: the mouth has to keep matching the words. Every take was filmed to playback and every take drifted slightly, so alignment comes before any creative decision about the cut.

1. Go to Vizard Agent and upload every take plus the song master.
2. Say the cuts should follow the beat without landing on every hit.
3. Say lip-sync must be preserved through the vocal delivery.

What do you need before you start?

The takes and the final master. Vizard Agent measures the drift between each take's on-set playback and the master itself, so the takes do not need to be pre-synced, but they do need to have been filmed to the same song rather than mimed to nothing.

• Every take. Wide, close, whichever you shot.
• The song master. The exact mix that will run underneath.
• Any B-roll. Short cutaways, filler, texture.
• The look. Grain, halation, contrast, whatever you want.
• The frame. Vertical for social, wide for a channel.

What do you type into Vizard Agent?

Name the two rules that govern a music video and leave the rest open. Vizard Agent needs to know that the beat drives the cut and that the lip-sync is protected, and once those are set it can be trusted with speed ramps, reverses and the grade.

Prompt

Variants worth knowing:

• Speed ramping and reverse. Used stylistically, off the vocal lines.
• Slow motion on the instrumental. Where no mouth is on screen.
• A specific film look. Contrast curve, grain, halation, split channels.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real vertical music video cut from five performance takes. The steps that decide whether the whole thing works come early: testing whether the clips were shot to playback at all, and then measuring each take's exact sync offset.

1. Probes every clip and the song, then downloads all the footage.
2. Builds contact sheets of every clip and reviews each take's frames.
3. Maps the song — the beat grid, the section structure and the energy.
4. Detects cuts inside each file and rebuilds the contact sheets correctly.
5. Tests whether the clips were shot to song playback, take by take.
6. Measures the exact playback sync offset per take, including the awkward fifth one.
7. Has each take watched shot by shot and analyses the track's hits and section energy.
8. Pulls word-level timings from the vocal so the cut can respect the phrasing.
9. Cuts to the beat grid, applies the grade and the speed work, and checks the lip-sync survived on the render.

Step 6 is the one that cannot be skipped. Two takes that both look in sync on their own will fight each other the moment you cut between them, and the fix is a measured offset per take rather than a nudge by eye.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 142.6 seconds, AAC audio. Vertical, two minutes and twenty-three seconds, cut from five performance takes plus short B-roll, graded with a filmic contrast curve and cutting on the song's structure rather than on every drum hit.

Two and a half minutes is a full song. The same method produces a shorter social cut from the same aligned takes, because the alignment work is done once and everything after it is selection.

When does this not work well?

Music video work fails in specific ways, and most of those failures are set on the shoot day rather than in the edit, which means no amount of care from Vizard Agent afterwards can recover them. These are the ones worth knowing before you upload.

• Takes not filmed to playback cannot be lip-synced. There is nothing to align to.
• A different mix breaks the sync. The master must match what played on set.
• Heavy speed ramps fight the mouth. They belong on instrumental sections.
• Five takes in one room still look like one room. Coverage is a shoot problem.
• Rights are yours. The song, the performers and the location are your responsibility.

How do you fix a result that came back wrong?

Say which section or which take. Vizard Agent keeps the beat grid, the per-take sync offsets, the contact sheets and the vocal word timings, so a chorus can be recut from different coverage without redoing the alignment work underneath it.

• "Use the close-up in the second chorus." Recut from the aligned takes.
• "It cuts too often." The beat grid is re-weighted toward section changes.
• "The grade is too heavy." Re-graded; the cut and the sync stay.

How does Vizard Agent compare to doing it yourself?

By hand this means syncing five takes to a master by waveform, checking each one holds through the whole song, marking the beat grid, then cutting. The syncing is the boring part and it is also the part that quietly ruins the edit when it is a few frames out.

Common questions

Do my takes need to be pre-synced? No. Vizard Agent measures each take's drift from the master and aligns them itself.

What if a take was not filmed to playback? Then Vizard Agent cannot lip-sync it. It can still be used on instrumental sections.

Will it cut on every drum hit? Not unless you ask. Vizard Agent cuts on kicks, snares and section changes by default.

Can Vizard Agent do speed ramps and reverses? Yes, and it places them off the vocal lines so the lip-sync is not disturbed.

How many takes should I shoot? Five gave Vizard Agent real choice here. Fewer works; the edit just has fewer places to go.

Why measure the sync per take instead of once? Because every take drifts by a different amount, and two takes that each look fine alone will visibly fight the moment the edit cuts between them. A measured offset per take is what makes the mouth hold across a cut.

Can I specify a film look? Yes. Vizard Agent applies contrast curves, grain, halation and channel effects as asked.

Can I get a vertical and a wide version? Yes. Vizard Agent re-frames from the same aligned takes without redoing the sync.

Does it use the vocal timing? Yes. Vizard Agent pulls word-level timings so the cut respects the phrasing.

Can I supply my own B-roll? Yes. Vizard Agent treats short cutaways and texture clips as free material because no mouth is on screen in them, which makes them useful for covering the joins between two takes that sit awkwardly together.

Does Vizard Agent check the finished video? Yes. It verifies on the render that the lip-sync survived the edit.