How to cut several people answering the same question without repeating

Last updated 2026-09-22 · 8 min read

What is the short version?

You filmed six people and asked them all the same five questions. Cutting the best answer from each of them in turn gives you six versions of the same answer, one after another. Vizard Agent compares across all the interviews before it chooses anything at all.

1. Tell Vizard Agent it must not get repetitive.
2. Let it group the answers by question, not by person.
3. Ask for the interviewer's questions to be cut out.

What do you need before you start?

The interview footage and the B-roll, and a note about which is which if the filenames do not say. Sampling every clip works and it costs a pass; telling it that four files are interviews and thirty are cutaways saves that.

Say whether the questions should be heard. Keeping them makes a longer, more conversational piece; cutting them makes a tighter montage where the answers have to stand on their own. Both are valid and they are different edits.

What do you type into Vizard Agent?

Name the failure you are actually worried about. "Make it engaging" is not something anyone can act on; "it shouldn't drag out or get repetitive" tells Vizard Agent precisely what to optimise against, and repetition, unlike engagement, is a measurable property of a set of transcripts.

Cut the interviews and B-roll together so each person shares their perspective, but it doesn't drag out or get repetitive. You can cross-cut two people on the same question.

That last permission is worth giving explicitly. Cross-cutting two answers on the same question is frequently better than choosing between them, and it is exactly the move Vizard Agent will not make unless you invite it.

What does Vizard Agent actually do?

It reads everybody before it cuts anybody. Each interview is transcribed with timecodes, the answers are separated from the interviewer's chatter, and the transcripts are organised by speaker so the same question can be compared across all of them at once.

In that session the work ran:

• Sampled the start, middle and end of every clip to separate interviews from usable B-roll.
• Transcribed the interview clips in parallel, keeping exact timecodes for clean cuts.
• Isolated the athletes' actual answers from the interviewer's talk.
• Organised the transcripts by speaker and clip so perspectives could be compared without everything going into the edit.
• Grouped consecutive phrases into complete sound bites rather than fragments.
• Read each interview looking for a distinct voice, explicitly avoiding a point the previous person had already made.

Then the review pass found repeated cutaways rather than repeated answers — the same shot used twice — and those were mapped back to their places in the edit and replaced individually.

What does the result look like?

A piece where each person on screen is there for a reason you could name. Nobody repeats what the person before them already said, the answers run as complete thoughts rather than fragments, and the B-roll covers the topic being discussed rather than simply filling time.

It is also usually shorter than you expected. Thirty answers reduce to a handful of distinct points, and the honest cut reflects that — which is why the brief has to say that length is not the goal.

When does this not work well?

When the repetition is itself the point. A testimonial reel where several customers independently say the same thing is using that agreement as its evidence, and de-duplicating it removes the entire argument. Tell Vizard Agent if that is what you are making.

It also struggles when everyone genuinely gave the same answer. If there are only three distinct perspectives among six people, then three is what you get back, and padding the piece out again means deliberately putting the repetition back in — which is a decision for you rather than something Vizard Agent should do quietly.

And cross-cutting needs audio that matches reasonably well to begin with. Two interviews recorded in different rooms cut back and forth badly however good the answers are, and levelling them helps only so far before the rooms themselves give it away.

How do you fix a result that came back wrong?

If someone you wanted in the piece is missing, say who. Vizard Agent will show you what that person's strongest distinct moment was and why it lost out, which usually turns out to mean their best answer duplicated somebody else's.

If the same shot appears twice, give both timestamps. That is the commonest repetition in a piece like this and it is a cutaway problem rather than an editorial one — Vizard Agent replaces the second instance without touching the answers.

If an answer starts awkwardly, ask Vizard Agent for a tighter start on that clip. Interview answers nearly always begin by restating the question back at the interviewer, and trimming into the substance is a per-clip fix that does not disturb anything around it.

How does Vizard Agent compare to doing it yourself?

An editor doing this properly makes a grid: questions down one side, speakers across the top, and the best answer in each cell. That is exactly the right method and almost nobody does it, because building it means watching six interviews with a notebook.

Reading the transcripts grouped by speaker builds that grid as a by-product. The selection is still a judgement, but it is made with all the answers to a question visible at once rather than in the order the tapes happened to play.

Common questions

Should I keep the questions in? Your choice. Vizard Agent can deliver both a tight montage and a longer conversational cut.

Can two people be cross-cut on one answer? Yes, and it often reads better. Say so and Vizard Agent will do it.

What if everyone said the same thing? Then Vizard Agent tells you, rather than including all of it.

Will each person definitely appear? Only if you ask Vizard Agent for that. Otherwise the strongest points win.

How does it find the answers? Vizard Agent separates the interviewee's speech from the interviewer's in the transcript.

Can I get the transcripts? Yes. Vizard Agent groups them by speaker, which is useful for the next edit too.

What about lower-thirds? Vizard Agent adds identifying names if you supply them.

How long should it be? As long as the distinct points support, and Vizard Agent reports how many it found.

Can it make a longer version too? Yes, with full question setups. It is built from the same selection.

Will the B-roll match the topic? Yes. Vizard Agent places cutaways over the answers they relate to.

Why did a shot repeat? Limited cutaway coverage. Vizard Agent flags repeats and swaps them.

Can I add more footage later? Yes. Vizard Agent places new material against the existing structure.

Does it fix uneven interview audio? It levels them, though very different rooms still cut noticeably.

Can it cut a version per person? Yes. Vizard Agent builds those from the same transcripts as separate deliverables.

Does the order of speakers matter? It does for the story. Vizard Agent sequences them by what each one adds rather than by who was filmed first.