The spine: The two characters, the setting and the silence at the end are fixed. The exchange is the variable.
Overheard works on recognition rather than information. Someone in your industry watches two characters say something they have heard fifty times, and the reward is the recognition itself — which is why the format travels through tagging rather than through saves.
Its defining constraint is that nothing is explained. No host, no framing, no lesson at the end. The gap between the two lines is the entire joke, and any attempt to close that gap with a caption saying “we've all been there” removes the reason it was funny.
Because it is two lines and one setting, it is the cheapest format on this list to produce at volume, and the only one on this page where a twice-weekly cadence is realistic without a team.
Best for: Brands whose audience shares an in-group vocabulary — agencies, trades, clinics, kitchens
The episode skeleton
Same beats every week. The timings are a starting shape, not a rule — what matters is the order and that no beat gets dropped.
- 10-1s
The setting card
Three words establishing where this is happening. It does the work a whole establishing shot would, and it is what tells the right audience this one is for them.
- 21-5s
Line one
The thing your customer, client or colleague actually says. Specific phrasing is everything — the more exactly it matches what people really say, the harder it lands with the people who have heard it.
- 35-10s
Line two
The reply. The joke lives in the distance between the two lines, so the second line should not comment on the first; it should simply follow it.
- 410-12s
The hold
Silence. No punchline card, no laugh, no explanation. The hold is what tells the audience the format trusts them, and it is the beat most brands cannot resist filling.
- 512-15s
The optional tag
At most a reaction — a look, a small gesture. Never a lesson, and never a product mention. A tag that sells cancels the recognition the previous ten seconds earned.
Decide once, then never again
This split is what makes a format cheaper every week instead of the same price every week.
Fixed forever
- The two characters and their locked references
- The setting and the three-word setting card
- The caption treatment for dialogue
- The silence at the end — the format is defined by what isn't there
Changes each episode
- The exchange, which is two lines and nothing else
10 episodes to start with
Adapt each one to your own category. At the stated cadence this is roughly two months of the series before you have to think of anything new.
- 01The question every customer asks in their first week
- 02What the client says versus what the client means
- 03The phrase a colleague uses that always means “no”
- 04The request that tells you a project is already in trouble
- 05Two specialists disagreeing about something tiny
- 06The onboarding call that goes sideways in a single line
- 07The thing everyone says at a trade show
- 08What a new hire says on day one
- 09The feedback line every person in your field has received
- 10The sentence said immediately before something goes wrong
How to build it in Versely
Every format on this hub is honest about how close the published library gets. This one has no workflow that runs in this shape — so the instruction is to clone the closest one and swap its asset map.
Nothing published is an overheard exchange. This is the closest structure — candid two-shot dialogue with four locked character references, handheld energy and native spoken dialogue rather than voiceover. Clone it and swap the asset map: keep the character locking and the two-shot framing, drop the interviewer, and let the two lines run without a question prompting them.
The pieces
Caption preset
InkWhite text on a solid dark card reads as film subtitling, which is the correct register for dialogue that is meant to feel observed rather than performed at the camera.
Once the first episode exists, the series itself is a schedule rather than a task: Run a recurring video series. Set it once. It writes a new version every time it runs.
Keeping it honest
Don't put a real, identifiable person's words in a character's mouth. Composite the exchange from things you've heard many times rather than quoting one customer, keep names, employers and identifying details out of it, and don't build the joke on a protected characteristic — the recognition should be about the situation, never about who the person is.
When to kill it
A format that has stopped working and keeps publishing is worse than no format at all. These are the signals to stop.
- You start explaining the joke on screen. That is the format asking to be retired, because the explanation means the lines stopped working.
- The lines stop being things people actually say and become things that would be funny if someone said them.
- You need three or four lines to land it. At that point it's a sketch, which is a different format with a different production cost.
Frequently asked questions
Do the characters need to be the same every time?+
Yes, and that is most of what makes it a series rather than a stream of jokes. Two locked characters in a fixed setting accumulate a relationship across episodes, and by episode ten the audience is bringing expectations to the exchange that you can play against.
How do I write lines that sound real?+
Write them down when you hear them, verbatim, including the clumsy phrasing. Real speech has redundancies and hesitations that invented dialogue smooths out, and the smoothing is exactly what makes written lines sound written.
Isn't this making fun of customers?+
It is if the customer is the target. The version that works has the situation as the target — both characters are recognisable, both are a bit ridiculous, and the person watching sees themselves in one of them without feeling looked down on. If you can't run an episode where your own side says the stupid line, the series will read as contempt.
Can I run it with generated characters?+
Yes, and it is the most practical way to produce it twice a week. Two locked character references plus dialogue is a small, repeatable job — the constraint is that the characters have to stay identical between episodes, which means reusing the saved assets rather than regenerating them.
Other formats worth running
Customer Question of the Week
The most durable format on this list, because the supply is your support inbox and it refills itself.
WeeklyThis or That
Two options your audience already argues about, one frame, and a verdict you're willing to defend.
Weekly, batching two or three submissions per episodeRate My X
The audience supplies the content and gets something useful back — the only format here where the supply problem solves itself.
Missing something you’d need to shoot this?
No camera, no studio, no presenter, no script — each gap has a route that still gets the episode made. The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.