How to find the clips worth posting in a stand-up set

Last updated 2026-08-24 · 7 min read

What is the short version?

Every comedian thinks they know which bits killed, and the recording disagrees more often than you would expect. Measuring the laughter after each line gives you an objective ranking, and then the job becomes framing rather than judgement.

1. Go to Vizard Agent and upload the full set.
2. Say you want short clips that work without context.
3. Say the length range and how many.

What do you need before you start?

The recording. Vizard Agent transcribes it, measures the response and finds the moments itself, so no marking is needed. Saying what you want the clips to do helps, because a bit that works for existing fans is a different clip from one aimed at strangers.

• The full set. Recorded from the room, with the audience audible.
• The length. Fifteen to forty-five seconds per clip.
• How many. Eight is a usable batch.
• What counts. A hook in the first seconds, a clean punchline, a real laugh.
• The shape. Vertical, with captions.

What do you type into Vizard Agent?

Name the criteria rather than the moments. Vizard Agent can measure a hook, a punchline and a laugh, so listing those as the selection rule gets you a ranked set rather than whichever bits happened to be near the start.

Prompt

Variants worth knowing:

• Clean only. Say so and Vizard Agent filters the material accordingly.
• A compilation as well. The individual clips plus one longer cut.
• Captions on. Comedy is watched muted more than comedians expect.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real stand-up set. The step that does the actual work is the nineteenth, and it is the reason this is not guesswork: the laughter after every line was measured.

1. Downloads the set and probes the source video.
2. Samples frames across the whole recording and reviews them.
3. Transcribes the set and logs its shots.
4. Reads the full transcript in three passes.
5. Maps when the comedian is actually in frame rather than when the camera is on the audience.
6. Builds a check grid of candidate moments and inspects the framing at each one.
7. Measures the audience laughter after every line across the whole set.
8. Checks the off-camera stretches so no clip starts on a cutaway.
9. Tracks his position and size in each candidate clip, locates his head with face detection, tests three crop framings for real and compares them, then renders all eight clips with styled captions.

Step 7 is the one worth stealing. Laughter is a measurable acoustic event, and sorting the set by how much of it followed each line surfaces bits the comedian had written off and demotes ones they were sure about.

What does the result look like?

From the run this page is written from, probed on one delivered clip: 1080x1920, H.264, 30fps, 18.23 seconds, AAC audio. Vertical, eighteen seconds, one of eight clips, framed with a tracking crop that follows the comedian across the stage, with styled captions burned in.

Eighteen seconds is short for stand-up and right for a feed. Vizard Agent cut to where the laugh peaks rather than letting it run, because a clip that continues past the response loses the energy that made it worth posting.

When does this not work well?

Laughter is a good signal and not a complete one, and a set recorded for archive purposes rarely frames well for a vertical feed. Vizard Agent measures the response and reframes the picture as far as the source allows, and both of those have real limits.

• A quiet room reads as a weak set. Laughter measurement depends on the audience being audible.
• Some bits need the room. A callback to something twenty minutes earlier will not work alone.
• The camera may be on the audience. Vizard Agent maps this so clips do not start on a cutaway.
• Comedy travels badly out of context. A line that lands live can read very differently as text on a feed.
• Venue and material rights. Where you recorded and what you perform may both carry conditions.

How do you fix a result that came back wrong?

Name the bit. Vizard Agent keeps the transcript, the laughter measurements, the framing map and the three tested crops, so re-cutting a clip or changing its framing is a re-render against analysis it already did across the set.

• "That one needs the setup before it." Extended from the transcript's word timings.
• "He drifts out of frame." Re-tracked from the face detection.
• "The bit at forty minutes is better." Checked against its measured laugh and cut.

How does Vizard Agent compare to doing it yourself?

By hand this means watching your own set back, which comedians famously hate, and choosing clips from memory of how the room felt. Then a fixed vertical crop puts you at the edge of frame every time you walk to the other side of the stage, which is most of the set.

Common questions

Does this work for other live performance? Yes. A lecture, a panel or a musical set all carry an audible audience response, and Vizard Agent can rank moments by it the same way. What changes is what counts as a reaction: laughter for comedy, applause elsewhere.

How does Vizard Agent know which jokes landed? Vizard Agent measures the audience laughter following each line and ranks the set by it, which turns the question into an acoustic fact rather than an opinion.

How many clips will I get? Eight from a set is a usable batch. Vizard Agent cuts what genuinely stands alone rather than filling a number.

Will I stay in frame? Yes. Vizard Agent tracks your position across the stage and builds a moving crop rather than fixing one in place.

Does Vizard Agent add captions? Yes, styled and burned in, generated from the set's own transcript rather than retyped.

Why does the laughter measurement matter so much? Because performers remember the bits they were nervous about and forget the ones that worked easily. A ranked measurement of the actual response regularly promotes material the comedian had written off, and it costs nothing to check.

Can Vizard Agent avoid clips that start on the audience? Yes. It maps when you are actually on camera before choosing any cut points.

Can I get a compilation too? Yes. Vizard Agent builds one alongside the individual clips from the same ranked selection.

Does Vizard Agent check the crops before rendering? Yes. It tested three framings on real frames on this run and compared them before choosing.