One concept.
Three ads that don’t look alike.
Three different executions of one idea, not one video exported three times.
The method is the machine. Our generator has years of our work inside it and writes every brief from that. No one here writes a word.
The concept is the idea. The ads are three ways of telling it.
A concept is a single argument about why someone should care about your product. One insight, one promise, one reason to stop scrolling.
The brief is built around the concept, not around a video, and no one here writes it. Our generator does. Years of doing this work by hand went into it, so it briefs the way we brief, every time, without a person in the loop. From the first line it already assumes the idea has to work three ways: what the creator has to record, which moments become the visual hooks, how the three cuts diverge. All of that is decided before anyone picks up a camera.
- A script.
- One script is one telling of the idea.
- A hook.
- A hook is the door. The concept is the room.
- Three exports of one video.
- That is a file, three times.
A concept is not
One exception: AIGC. There the three ads are deliberately the exact same ad, with the same script, hook and shots, performed by three different people. The person is the only thing that changes, so you learn which face works.
Since 2026, this stopped being a nice to have.
Meta’s Andromeda system groups ads that look alike under a shared Entity ID. Two ads with the same footage and a different headline are not two ads to Meta. They are one.
And when most of your library looks the same, a large part of it never gets past the retrieval stage, which means it never even enters the auction. Your variation dies before a single person could have seen it.
Ten visually distinct ads give the platform ten chances to match you to someone. Ten near identical ones give it one.
56%
of digital sales lift comes down to the creative itself
Nielsen
3×
higher purchase intent running 10+ distinct creatives vs fewer than five
TikTok
5 to 7 days
for a concept to fatigue now, down from 14 to 21
Meta, post Andromeda
~50%
of all creatives are retired before day 28
Motion, across $1.3B of spend
Same concept. Five deliberate differences.
The visual hook
Always different. Never optional.
The first frame is the autoplay thumbnail, and it is also the clearest signal of whether two ads are the same ad. Change it and the platform reads three creatives instead of one. Above a 25% hook rate is table stakes. 40% is where the algorithm starts rewarding you with cheaper reach. Opening on a face with direct eye contact alone moves it 4 to 8 points.
The caption hook
Three points of view on one idea.
Same concept, argued three ways. And the caption has to agree with what is on screen: a caption promising one thing over footage showing another is the fastest way to lose someone in the first second.
The talking head
Same recording. Different telling.
Lines are cut or reordered so the story lands at three different speeds. Some people want the point immediately. Some need to trust the person before they will hear an offer. Some are not listening at all and are only watching. One pacing cannot serve all three.
The b-roll order
Same footage. Different sequence.
Which moment leads changes what the ad is about. Move the result to the front and it is a proof ad. Move the problem to the front and it is a pain ad. Same clips.
The editing style
Organic, polished, interface led.
One cut stays raw and native. One is polished, with motion and graphics. One puts the product on screen and makes the creator the supporting act, the inverse of the first. Native feeling content sees around 27% higher completion, but raw is not automatically better: the authenticity that performs is in how someone talks, not in how little you edited. So we test the axis properly instead of assuming.
Changing one thing is not a variation. Changing the ad without changing the concept is.
Swapping a headline, trimming two seconds, adding a different end card. The platform sees straight through all of it, and so does your audience. A real variation has a different first frame, a different opening line, a different rhythm and a different edit, and still argues exactly the same thing. That is harder to make. It is the whole job.
Three executions tell you which axis your audience responds to.
When three ads share a concept but differ on purpose, the winner tells you something you can use again. It was not luck. It was the pacing, or the style, or the angle.
Run enough concepts this way and you stop guessing what to make next: you know whether your buyers want proof or story, polish or raw, product or person. That knowledge compounds. A single winning ad does not.
Brief, creators, edit. Planned as three from the start.
The thinking is the product. Our generator holds years of client work, what we have briefed, what ran, what won, and it writes every brief from that, the way we would. Nobody edits its output into shape and nobody writes alongside it.
What people do is execute. Creators from our community of 1,000+ record the material the brief asks for. Our editors cut the three versions to the styles it specifies. Skilled work, but it is the brief that decides what the three ads are.
How it works