Building AI ads like a scientist🧑🏻‍🔬

The hardest ads to make right now aren’t the ones that work. It’s the ones that work again after your last winner dies. That’s why we build moonshots - not because it’s trendy, but because it’s the only way to stay ahead of fatigue math. And AI isn’t a gimmick here. It’s a weapon for speed-testing what actually deserves spend.

Previously, we showed you how we use “moonshots” to fight creative fatigue. When tweaks, edits, and remixes stop working, the only path forward is to build something entirely new.

Today, we’re taking you inside one of those “moonshots”: How we actually create high-quality AI-powered video ads at Inceptly.

🎭 The Nerd vs. The Rocker

You’ll find yourself in one of these two people. And if you’re lucky? You’ll become a little bit of both.

The Nerd craves structure. Spreadsheets. Research. Precise control over every creative detail.

The Rocker? Operates on instinct and momentum. He needs to see results fast, or his energy fades.

Let’s follow them both and see how each builds an AI video from scratch.

🧠 The Nerd's process: From reference boards to render queue

First, the Nerd defines the concept.

What type of video is this? Who are the characters? What’s the setting? He needs to see every detail from different angles, in different contexts.

So he opens Pinterest, Sora, Midjourney - maybe all of them.
He builds reference boards. He studies. He tweaks.

Once he’s clear on the vision, he writes the script.

Then he breaks the script into short scenes and phrases.

Only then does he move on to creating the images that will serve as the storyboard for the future video.

His first stop? Character generation.

Not ChatGPT (anymore). The Nerd now uses NanoBanana, a new AI image model that’s exploded across Twitter.

Why? Because its photorealism is scary good. Even the Nerd couldn’t tell the difference.

Input:

Output:

He creates the characters, then sends them back into NanoBanana - this time, placing them into specific scenes and environments.

Whatever fits the vision goes into the storyboard.

With visuals locked, it’s time to animate.

• If he’s got budget: Google Veo 3 for instant image-to-video with voiceover.

• If not: Runway ML for animation only, plus AI lip-sync tools for voice.

What’s the difference?

Google Veo 3 lets you instantly animate and voice your image.

Runway only animates, and not even on the first try. You’ll need to add voice later using an additional tool for Runway - AI lip-sync.

Once all video clips are ready, the Nerd opens Premiere Pro or DaVinci Resolve.

He assembles the footage, adds music, sound design, and typography.

(And now he’s ready to argue with media buyers about those “one last round” edits, but that’s another story.)

And here is the result - the full AI-animated “moonshot” video we made for our client:

The Nerd shows how AI ads get built with precision - every frame planned, every scene mapped.

But sometimes? Speed beats structure.

Next week → The Rocker’s flow

We’ll walk through the instinct-first path to AI ads: fast, raw, and shipped before fatigue catches up.

Stay tuned,

The Inceptly Team

💌 Like this newsletter? Let's continue the conversation

Get in touch with us by responding to this email or tagging us on LinkedIn or Instagram and sharing your thoughts. Your feedback helps us keep our newsletter relevant and interesting.