One brief, hundreds of ads
A single brief: one product, one core claim, one feeling you want the viewer left with. From that, an AI production pipeline can make hundreds of ad variants, with different people, different openings, different settings, different languages and different cuts.
That changes the economics of testing. With creators, every new face is another person to brief, pay and ship product to, so most brands test two or three and pad the account with hook variations. With AI UGC, a new face is a production decision. Our pricing puts a finished ad video at ₹3,500 to ₹6,500 on a 30-video month, while experienced creators in India commonly charge several times that before usage rights. We break that down in what a UGC video really costs in India.
Production stops being the constraint. Pipeline design becomes it.
The brief comes first
A pipeline is only as good as its brief, and ours is a structured spec, not a mood board:
- The product, and the one benefit that matters, in a sentence
- The customer's problem, in their own words
- The feeling the ad should leave
- The claims we're allowed to make, and the ones we aren't
- The call to action
Every variant inherits from this. A vague brief gives you hundreds of variants that average out to the same generic ad.
The variables: who, what, where, and how it's cut
From the brief, we branch on four variables:
- Persona: who's talking. Age, gender, region, language, life stage.
- Angle: what they say. Problem first, result first, a switch from another product, an objection, a routine.
- Setting: where it happens. Bathroom mirror, office desk, gym, kitchen, a street outside.
- Format and cut: talking head, hands-only demo, before and after, montage, in 9:16, 4:5 and 1:1.
Five personas, five angles and four settings is already 100 concepts, before you cut a single aspect ratio.
The order matters more than it used to. Since Meta's Andromeda update, ads that look alike tend to compete with each other, so we vary persona and setting first and treat new hooks on the same footage as edits, not new ads. The reasoning is in our Andromeda playbook.
The production stack
Images. Gemini and Flux image models make the base frames: the person, the product in hand, the setting. The real product photo goes into every generation as a reference image, so the label and shape stay accurate. Prompts describe the feeling in the brief, not just the object. "A woman in her thirties at a bathroom mirror, morning window light, holding the serum, candid phone-camera feel" is a prompt. "Skincare product" isn't.
Motion. Kling turns approved frames into clips of 5 to 10 seconds, with one camera move at a time, because compound motion is where faces start to morph.
Voice. ElevenLabs for voiceover in the right tone and language.
Edit. Captions, timing and music, by hand or programmatically with Remotion once the volume justifies it.
Three review gates
Volume without review is a liability. Before anything reaches a media buyer, it passes three gates.
Gate one: rules in the prompts. Brand constraints live inside the generation prompts: the correct product colour, approved claims only, no invented text on the packaging. This stops the obvious mistakes before they're made.
Gate two: a person approves the concepts. Nobody reviews hundreds of videos one by one. A person reviews the 10 to 20 concepts the variants come from, plus a sample of finished videos from each. If the concept and the sample pass, the variants carry that approval.
Gate three: platform and compliance. Meta labels some AI-generated or AI-edited ads, and beauty and health claims have to stay inside ASCI's rules. This is a checklist at the end, not a creative step.
What it has produced
We've used this approach to deliver more than 1,500 ads for Adkrity and Xyxx, and more than 1,000 multilingual content pieces for a group of beauty brands: Mamaearth, Aqualogica, The Derma Co, Dr. Sheth's and Dr. Facts. The case studies have the details.
What changes for creative teams
The job moves from making each ad to designing the system that makes them: writing briefs that hold up across hundreds of variants, keeping a library of angles that work, maintaining the prompt templates, and running the review gates.
That's a smaller team doing more valuable work. If you'd like to see it on your own product first, we'll make a free sample ad within 24 hours.
