B-roll in an ad is the picture that proves the sentence

You are filming one main shot, usually a person speaking to camera. Everything you cut in over the top of that shot, while their voice continues, is b-roll.

The continuous audio is the whole test. If the sound changes when the picture changes, you have made a scene. A scene has to match performance and continuity as well as picture. That is a shoot day. A cutaway is two seconds of footage you already have.

In a paid ad the cutaway does a second job that has nothing to do with information. It resets attention. A picture that has not changed for eight seconds is a picture the viewer has finished looking at.

So b-roll in an ad carries two loads at once. It proves the sentence. It buys the next three seconds.

That is why a talking-head ad with no cutaways underperforms one with them. Better camera work does not close the gap. It is not a quality problem. Nothing on screen changed.

One more consequence, worth knowing before you place anything. A cutaway also hides a cut. Delete a sentence from the middle of a take and the speaker's head jumps. A clip over the join makes the removal invisible.

Thirty seconds for a clamp-on desk lamp, filmed by the person who bought it, in the room where it now lives.

Second 0 to 4. Face to camera. He says he was getting headaches at about four in the afternoon and blamed his monitor. Main shot, no cutaway, because the claim needs a face behind it.

Second 4 to 8. Cutaway to the ceiling light in his room, a bare bulb in a bad fitting. He keeps talking. The picture does the accusing so his sentence does not have to.

Second 8 to 12. Cutaway to hands clamping the lamp to the desk edge, one turn of the screw. People copy this one wrong. It is not showing the product. It is showing that installing it took four seconds.

Second 12 to 17. Back to his face for the price and the reason he almost did not buy it. Never cover the moment the claim lands. The viewer wants to see who is saying it.

Second 17 to 24. Cutaway to the pool of light on the keyboard, then the same desk with the ceiling light off. The proof shot. It is the longest hold in the ad.

Second 24 to 30. Face, the offer, done. Total coverage is fourteen seconds of thirty, a shade under half.

Placement is where most ads go wrong. The rule is simpler than people expect. Cover the words that name something you can photograph.

A sentence with a concrete noun in it has a picture waiting. The collar, the box, the screw, the keyboard, the empty bottle. A cutaway there reads as evidence.

A sentence about a feeling, a guarantee or a comparison has no picture. Put footage over it and the ad reads as a stock montage with narration. The viewer can tell the clip was chosen to fill time.

Three moments stay uncovered no matter how good the footage is.

The first sentence, because the viewer is deciding whether a person is talking to them. The moment the claim lands, because a claim needs a face behind it. The price, because covering a number reads as hiding it.

Everything else is available. In practice about half the ad is coverable, and you should use roughly all of it.

There is a point where cutaways stop helping and start replacing the ad. Past roughly half the runtime the viewer is no longer watching a person talk. They are watching footage with a voice on it, and the trust the format was built on is gone.

So the ceiling is enforced rather than recommended. At chill pace no more than 40 percent of the runtime can be covered. At normal pace 45 percent. At fast pace 52 percent.

Those are compiler constants, not advice in a help article. Highlight more phrases than the cap allows and the cap still holds.

The number moves with pace for a reason. A faster read has shorter sentences and more of them, so there are more places a picture can change without the ad feeling restless.

You will want more coverage than is good for you. It happens on the third or fourth ad of a session, when the take starts to feel repetitive. It does not feel repetitive to somebody seeing it once.

The ceiling protects the ad from that instinct, on every export, without you having to remember it.

The old loop is the reason most ads have no b-roll in them. Watch the take. Note the phrases that need a picture. Open a stock library. Search, preview, download, import, drag onto a second track, trim to the sentence, nudge, watch it back, nudge again.

Forty minutes per ad. The twelfth ad in a month gets none of it, because there is no forty minutes left.

Here the sourcing and the first placement happen inside the batch. One take of up to three minutes goes up for 100 credits. The cut comes back with footage already sitting on the phrases that had concrete nouns in them.

It gets some of them wrong, which is expected on a first draft. You fix it by highlighting a different phrase, and a clip lands over exactly those words. Swap the clip in any spot, or search millions of free ones.

Change how each clip enters if you want. Cut, Whip, Punch, Glitch or Sweep. Set the pace and the whole thing re-cuts inside the ceiling.

Export is 20 credits per output minute, so a thirty second ad costs 10. The trial is 7 days and 300 credits with no card. That is two takes and four thirty second ads with the coverage already placed.