How to cut a reel around people's reactions

Last updated 2026-08-25 · 7 min read

What is the short version?

A reaction reel is built on one moment: the point where things change. Everything before it sets up the mood and everything after it is people's faces, so the edit is really a matter of finding the turn and then finding the best expressions around it.

1. Go to Vizard Agent and upload the footage.
2. Say where the mood turns and what it turns into.
3. Ask for the reaction faces, zoomed and held.

What do you need before you start?

The raw footage and a description of the turn. Vizard Agent finds the faces and the dialogue itself, so timestamps are unnecessary, but describing the moment in plain words is what tells it which part of a long recording matters.

• The footage. Uncut is fine.
• The turn. What happens, and roughly when.
• The length. Forty-five seconds is a good reel.
• The subtitle style. Fast-paced, bilingual, plain.
• The tone. Funny, dramatic, or straight.

What do you type into Vizard Agent?

Describe the arc rather than the timestamps. Vizard Agent matches your description against the transcript and the visual shot log to locate the moment, which is far more reliable than you scrubbing through a long recording trying to read a timecode off the screen.

Prompt

Variants worth knowing:

• Bilingual subtitles. Two languages, cut fast.
• Zoom held on a face. Rather than cutting away from it.
• Reaction grids. So you can see the candidate expressions.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real reaction reel cut from one long clip. The step that makes the reel funny rather than merely coherent is the face scan: it finds every face in the footage and lays the expressions out in a grid to be compared.

1. Inspects the uploaded video and transcribes the dialogue.
2. Analyses the shots across the whole recording and extracts frames from the key moments.
3. Builds visual contact sheets and reviews both halves of the arc — the setup and the turn.
4. Scans the video for face appearances and extracts every face and reaction frame.
5. Reviews the reaction frames in a grid, in two passes, to compare the expressions.
6. Pulls precise word timings for the setup line and the turn.
7. Extracts frames from the turn sequence specifically and reviews them closely.
8. Builds a grid of the best reaction frames and picks the ones that carry the moment.
9. Checks the vertical crop positions on a preview grid, then cuts, zooms and subtitles the reel.

Step 5 is why the reel lands. An expression that seems perfect while you are scrubbing past it often looks ordinary next to five others, and the only way to know is to put them side by side.

What does the result look like?

A vertical reel of around forty-five seconds: an easy, cheerful opening that establishes the mood, a hard turn at the moment everything changes, and then a run of held close-ups on faces with fast subtitles carrying the dialogue in both languages.

The reactions are zoomed in from the original frame rather than shot that way, which is what makes this possible at all — the footage was never filmed with close-ups, and the crop supplies them.

When does this not work well?

Reaction reels depend entirely on what the camera happened to catch on the day, and no amount of editing by Vizard Agent supplies a face that was pointed the wrong way or a turn that only exists in the audio.

• Faces have to be in frame. A camera on the road catches no expressions.
• Zooming costs resolution. A soft wide shot gets softer when cropped in.
• The turn has to be visible. If it only happens in the audio, there is nothing to cut to.
• People are identifiable. Posting strangers' reactions has consequences.
• Funny is subjective. Vizard Agent picks the strongest expressions, not the funniest to you.

How do you fix a result that came back wrong?

Point at the face or the beat. Vizard Agent keeps the face scan, every reaction frame, the word timings and the crop previews, so a different expression can be swapped in or the turn re-timed without recutting the whole reel.

• "Use his face, not hers." Swapped from the reaction grid already built.
• "The turn comes too late." Re-timed against the word timings.
• "The crop cuts off his head." Repositioned from the crop preview grid.

How does Vizard Agent compare to doing it yourself?

By hand this means scrubbing frame by frame looking for the good faces, which is slow and unreliable because you cannot hold six expressions in your head at once. Then each one has to be cropped and keyframed into a vertical frame separately.

Common questions

Do I need to give timestamps? No. Describe the moment and Vizard Agent locates it from the transcript and the shot log.

How does Vizard Agent find the faces? It scans the footage for face appearances and extracts every reaction frame it finds.

Can it zoom in on expressions? Yes. Vizard Agent crops in from the original frame and holds on the face.

What resolution do I need? High enough to survive the crop. Vizard Agent zooms into the source, so softness magnifies.

Can the subtitles be bilingual? Yes. Vizard Agent cut fast Hindi and English subtitles on this run.

Why compare the faces in a grid? Because an expression that looks perfect while you scrub past it often turns out ordinary next to five others. Laying the candidates side by side is the only way to pick the one that actually carries the moment.

How long should a reaction reel be? Around forty-five seconds. Long enough to set the mood, short enough to stay on the turn.

Can I use footage filmed horizontally? Yes. Vizard Agent checks the vertical crop positions on a preview grid before committing.

What about people who did not agree to appear? That is your call to make. Posting identifiable strangers carries real consequences.

Can I get several reels from one recording? Yes. Vizard Agent keeps the face scan and the transcript for the whole video, so a second turn elsewhere in the footage can be cut without repeating any of the analysis work.

Should the reel open on the turn or build to it? Build to it. Vizard Agent uses the cheerful opening to establish what is about to be broken, and a reel that opens on the panic has nothing to contrast it against, which is why the setup earns its seconds.

Does Vizard Agent check the finished reel? Yes. It reviews the crops and the subtitle timing before delivery.