Lesson 5 — The Video Assembly Line
One winning ad is luck. A system that ships twenty variants a week is a business. Build the line, not the video.
- Treat hook, body and CTA as swappable modules: 5 x 2 x 2 yields twenty ads from nine generated pieces.
- Draft in a cheap, fast tier and spend the premium tier only on the concepts that survive the cut.
- Track cost per finished ad including failed rerolls; AI variants are usually far cheaper than a crew shoot, but only if you count the waste.
- Name files so every winner traces back to its exact hook, body, CTA and version, then run a fixed weekly ship-and-read cadence.
Build ads as three swappable modules: hook, body, call to action. Five hooks, two bodies and two CTAs give you twenty finished ads from nine generated pieces. You recombine instead of rebuilding, and volume stops being expensive.
Generation is priced per second, so draft everything in the cheap fast tier first. You can judge framing, motion and hook strength at draft quality. Only the concepts that survive that cut earn the expensive final render.
Track cost per finished ad, and include failed rerolls, because that is where budgets disappear. AI variant testing generally costs a small fraction of a live crew shoot, but only if you count the waste instead of ignoring it.
Name every file so it identifies the product, the hook number, the body, the CTA, the ratio and the version. When a variant wins, you need to know exactly which module did it. Untraceable winners cannot be repeated.
Run a weekly rhythm. Write hooks, generate, voice and cut, ship, then read the data and kill the losers. Next week's hooks are variations of this week's winners. The system compounds; a one-off viral video does not.
Action step. Pick one product and build a grid of five hooks, two bodies and two calls to action. Draft everything in the cheap tier, finalize the survivors, name every file properly, ship twenty variants and log the results.
Loading the quiz…