How We Made the Better AI Video
Most people who see an AI video like this ask "what prompt did you use?" — but the prompt is not where the video starts. The Better video started with one question: how do you make someone fly from New York to Israel for a product that only exists there? The moment that question existed, there was a story — and from then on the prompt is just a record of directing decisions already made.
1. We did not start from the product
The brief could easily have become an ordinary product video: here is the box, here are the products, here are the benefits, here is the call to buy. The problem is that a video like that may show the product, but it gives nobody a reason to stop scrolling.
So we started from an exaggerated behaviour the viewer wants to understand: Monica sees the product on her phone, discovers it is only sold in Israel, and decides to get there immediately.
A question
Why does she want this product so badly? The viewer stays to find out.
A conflict
Will she even manage to get it? There is something to lose.
A promise
We will watch her reach the shop and fight for it. There is somewhere to arrive.
If the idea does not produce a question, a conflict and a promise in its first three sentences, no prompt will save it.
2. The hook: from calm to crisis
In the first version the hook opened too calmly. Monica saw an ad on her phone, but it was not immediately clear why it mattered. We sharpened the opening into a moment of crisis:
"Oh no! Better is only in Israel?"
That line works because it does three things at once: it carries emotion, it carries information, and it pushes the story forward.
The order of the frames is a decision too. The first frame shows Monica sitting on a sofa — her face clear, her Star of David clear — and only then does the camera move to what she is seeing on her phone. The viewer connects to the character first, and understands what she saw second. Reverse that order and you have an advert.
3. A short story track, not one long video
Instead of writing one continuous video, we broke it into distinct beats. Each beat has to answer one question or create a new one — and if it has no job, it probably does not belong in the video.
The discovery
Monica finds out the product is only available in Israel.
The call
She calls her agent.
The decision
She books a flight.
The landing
She arrives in Israel.
The chase
She runs to Super-Pharm.
The find
She finds the last box on the shelf.
The obstacle
An Israeli customer arrives at exactly the same moment.
The stand-off
A comic tug-of-war over the box begins.
The information
The product benefits enter as part of the argument.
The ending
The product becomes a moment of identity and feeling.
4. Why we split the video into two prompts
A video like this involves several locations, changes of pace, dialogue, a product, a lead character and a second character. Put all of it into one prompt and the model has to solve too many problems at once — which shows up as jumps in location, a changing face, dialogue out of sync, or a product that warps.
Part one
Builds the motivation and gets Monica all the way to Super-Pharm.
Part two
Opens at the shelf and focuses on the conflict, the product and the ending.
It is a directing decision. Each part carries exactly one central idea, which is why the model manages to hold it.
5. References assigned by role, not by quantity
We did not upload every image we had. Each reference was given one job and one job only:
The principle is simple: a character reference should not govern the product, and a product reference should not alter the character. The clearer each image's job, the less the model confuses the things it is meant to preserve.
6. The benefits entered through the conflict
We did not want to stop the story so somebody could deliver a marketing line. So Better's benefits entered as part of the argument between the two women:
"Better absorption. Up to five times more."
"Better experience. Gummies you actually want to take every day."
That way the information does not feel like a sales slide. Each of them is trying to prove why she needs the product more, and along the way we understand what makes it interesting. That is the difference between putting information into a video and making the information part of the story.
7. Protecting the product inside the AI
This was one of the biggest challenges. The moment two characters pull at a box, the model may let the products fall out of it, change shape, disappear, or turn into something else.
So "keep the products in the box" was not enough. We defined physical rules:
- The box and its three products are locked objects.
- Both characters may touch only the outer sides of the box.
- They never touch the products themselves.
- The box moves as a single unit — the products stay in their original positions.
That is the kind of precision that turns a nice idea into a result you can actually put in front of a client.
Want the full process, not just the prompt? Open the directing guide in OMER METHOD →
What you do not see in the video
You do not see the attempts that failed. You do not see the versions where the hook was too weak. You do not see the products that changed colour or fell out of the box. You do not see the decision to split the story into two prompts. And you do not see all the times you have to stop and ask: what is the viewer supposed to feel right now?
That is the real work in making video with AI — not just making a frame look good, but making every moment serve the story.
Seven questions to answer before you open an AI tool
- What is the idea the viewer must grasp in the first few seconds?
- What does the character want?
- What stands in her way?
- Which information can become part of the action instead of being recited?
- Which reference governs which detail?
- Where might the model fail physically or visually?
- What is the last frame the viewer should remember?
If you do not have clear answers to those seven questions, another prompt probably will not solve the problem.
The secret is not the prompt
This video did not happen because we found one magic sentence. It was built through a process: idea, hook, story, breakdown into scenes, references, directing, failure checks and fixes.
AI is a very strong production tool. But it does not decide for you why anyone should care. The same principles sit behind everything I build — from Monica Andrew to the CRINIÈRE production file. And that is exactly why I built OMER METHOD.
Want to learn the whole process?
In the guide library I break down how an early idea becomes an AI video people stop for, watch to the end and come back to. Not just how to write a prompt — how to crack an idea, build a story, direct a character, work properly with references and hold a consistent result across the whole video.