The finished ad, and the 3-step system | Academy

Every frame in this commercial — the kitchen, the stadium, the street, the office — was generated, on a laptop, with no camera and no shoot.

Watch the full spot above. Everything below breaks down exactly how it was built, so you can run the same system on your own product, brand, or idea.

The system is three steps, not one long prompt

There's no single magic prompt behind a commercial like this — it's a pipeline. Skip a step and the rest falls apart: no locked assets means the model hallucinates a different face every scene, and no shot list means Seedance has nothing structured to animate.

Step	What you build	Where
1. Assets	Product, hero, locations, props	Soul Cinema + GPT Image 2.0
2. Shot list	A Claude skill turns the story into structured prompts	Claude
3. Scenes	Every shot generated and assembled	Seedance 2.0, inside Higgsfield AI

Lock this order in your head before you open anything: assets first, prompts second, generation last. Generating scenes before your assets are locked means nothing stays consistent from shot to shot — the model has nothing fixed to hold onto.

Step one is where the real work lives, and where most of the consistency problems get solved before they ever reach a scene. That's next.