How to cut interview clips for the guest rather than the show

Last updated 2026-09-22 · 8 min read

What is the short version?

The same interview yields two completely different sets of clips depending on who is posting them. The show's set is about the conversation being worth watching; the guest's set is about the guest being worth following, and Vizard Agent selects against whichever standard you name.

1. Tell Vizard Agent whose social media the clips are for.
2. Say what expertise the guest should come across with.
3. Ask for moments that stand alone without the question.

What do you need before you start?

The recording and a sentence about the guest's positioning. Vizard Agent reviews the whole interview, but knowing that the guest wants to be known for second homes and rentals rather than for the market in general changes which fifteen moments matter.

• The full interview. All of it, not the highlights.
• Whose feed. The guest's, the show's, or both.
• What they want to be known for. The specific niche.
• How many clips. Ten to fifteen is a normal ask.
• A reference for the look. Colours, hook style, captions.

What do you type into Vizard Agent?

Say who the audience is and what they should think afterwards. A clip that makes the guest look knowledgeable is a different selection from a clip that makes the conversation look lively, and only you know which one you are paying for.

Prompt

Variants worth knowing:

• "For the guest's own social media." The instruction that changes everything.
• "Work without the host's question." The selection standard.
• "Keep him in frame rather than the host." The crop rule.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a 53-minute interview cut into fourteen clips for the guest's own channels. The selection standard is written down and applied explicitly rather than left as a matter of taste, which is what makes the batch consistent.

1. Looks at the hook reference and matches its colour treatment.
2. Builds a word-timed transcript so clips can start and end on exact words.
3. Reviews all 53 minutes, not only the opening, for distinct standalone moments.
4. Checks the camera layout so the vertical crop keeps the guest, not the host, as the focus.
5. Separates practical advice from introductions and tangents.
6. Judges each passage against one standard: useful without the host's question.
7. Screens the middle of the interview for the specific subjects the guest wants to own.
8. Refines exact sentence boundaries and checks the guest's name and company spelling.
9. Removes the host's unnecessary parts from each selected passage.
10. Tightens pauses and repeated starts without flattening the guest's natural delivery.
11. Favours evergreen lessons over dated price claims, keeping the context around them.
12. Checks the cuts word by word so no clip starts on a fragment or ends on the host.
13. Tests the vertical crop across all fourteen selections before rendering any of them.
14. Tests caption sizing on one rendered sample before applying it to the batch.

Step six is the whole method. "Would this make sense to someone who has never heard of the show?" is a test you can apply to a passage in seconds, and it eliminates most of what feels good while you are watching the interview.

Step eleven is the difference between clips that work for a year and clips that are embarrassing in March. A number that was true in the recording becomes a claim when it is posted six months later, so Vizard Agent prefers the lesson to the figure.

Step thirteen saves the batch. Testing the crop on one clip proves nothing when the guest moves in his chair over 53 minutes; testing all fourteen catches the two where the framing fails.

What does the result look like?

Fourteen vertical clips that each open on a hook, stand alone without the interviewer, keep the guest in frame and in focus, start and end on whole sentences, carry consistent captions and identification, and say things that will still be true next year.

When does this not work well?

Some interviews will not yield guest-first clips. If the host talks as much as the guest, if the best material is a back-and-forth, or if the guest was only ever framed in a two-shot, the clips end up being about the conversation whatever the brief says.

• Genuine dialogue. The value is in the exchange.
• A two-shot only. No clean crop to the guest.
• A guest who only answers. No standalone statements.
• Heavily dated material. Prices, rates, current events.
• Poor audio on one side. The guest's track is what matters here.

How do you fix a result that came back wrong?

Say which clip it is and whether the problem is the selection or the framing. Vizard Agent treats those as two separate stages, so a note about the crop on one clip does not send the whole batch back through the selection process again.

• "This one needs the question to make sense." It is dropped or the setup is included.
• "The host is in shot." That clip's crop is rebuilt around the guest.
• "It starts mid-sentence." The boundary is moved to the word before.
• "That price will age badly." The passage is swapped for an evergreen one.

How does Vizard Agent compare to doing it yourself?

By hand you scrub for good moments, and "good moment" quietly means "moment I enjoyed", which is usually a laugh that needs the question, the host and the context. Then you crop to the middle of the frame and the guest is half out of it.

Common questions

Can it do both sets — guest and show? Yes. Tell Vizard Agent and it selects twice against two different standards.

How many clips from an hour? Ten to fifteen strong ones is realistic; Vizard Agent will say if there are fewer.

Will it keep the guest's voice natural? Yes. Vizard Agent tightens pauses without cutting the rhythm of their speech.

Can it add the guest's branding? Yes — name, title and company, spelled as Vizard Agent verified them.

What if the host asks a great question? Then include it, and Vizard Agent keeps the exchange as one clip.

Can it match a hook style I like? Yes. Give Vizard Agent the reference and it matches the treatment.

Does it caption them? Yes, in one consistent layout that Vizard Agent tests on a sample first.

Can it avoid dated claims? Yes, that is a standing instruction you can give Vizard Agent.

What about clips for the host instead? Same process, opposite crop and a different selection standard.

Can it produce a posting order? Yes. Ask Vizard Agent to rank the clips by strength.

Do the clips need the show's logo? Usually not on the guest's feed, but that is your call.

How long should each clip be? Whatever the thought takes — Vizard Agent cuts to the sentence, not the second.

Can it do horizontal versions too? Yes, from the same selections, for the show's own channel.

What if the guest misspeaks? Vizard Agent flags it rather than posting it — you decide whether to keep it.