Hey Party People,

Quick life update before we get into it: it's Linus's birthday this week. He turns 5! We got him a cake. We'll see if he actually eats it or just stares at it. Forgotten first “born” child. He almost caught a squirrel today, and gave two people heart attacks. So that was exciting. 🙃

James is currently obsessed with a busy board, like genuinely locked in, building that work ethic from day 1, son. Like you put it in front of him and all bets are off. My man is flicking and twirling like it is his job.

Meanwhile, I'm writing this at 9pm Thursday because procrastination is my brand.

Maddie and I have date night at Swoonie's tomorrow night (Brooklyn Heights (I think / ish), highly recommend, low-key vibes).

For Halloween, James is dressing as a Grateful Dead dancing bear. Yes, we're those people. We walked down the aisle to Good Lovin'. Our WiFi password is a Dead lyric. Don't @ me. Anyway. Q4 chaos. You get it.

Now, let's talk about AI creative production.

So.
My DMs are full.

Not in the fun influencer way.
In the "Everyone wants AI consulting and I'm tired of saying the same thing on Zoom" way.

So here's what we're doing instead: For the next 5 weeks, I'm going to show you exactly how to use AI for creative production. One workflow per week. One email. No fluff.

By Week 5, you'll have a production system your competitors don't. Or you'll have ignored all of this and still be using ChatGPT to write emails slightly faster. Your choice.

Here's what we're covering. One week. One workflow. One skill.

Week 1 (This Week): Static to Motion in Midjourney
Take your proven statics, make them move, feed them to Andromeda. 10 ads become 40+.

Week 2: Fixing AI Weirdness with Seedream
Your AI outputs look plastic. Here's how to make them not suck.

Week 3: Concept Testing with Sora
Kill bad creative ideas before you waste $30K shooting them.

Week 4: Product Insertion with Nano Banana
Put your product anywhere. (With very reasonable expectations.)

Week 5: Full B-Roll Production with Veo 3.1
Everything combined. 28-minute sequences. This is the final boss. Think Ganondorf in Ocarina.

Each week builds on the last. Don't skip ahead. I know product insertion sounds sexier. You're wrong. Trust me.

Start here:

This is where everyone should start. Not product insertion. Not Sora. Here.

Why?

Because you already have assets that work. You know they perform. You have data. You have statics that are PROFITABLE.

Now you're going to make them move.
- Generate 3-5 motion variations per static.
- Test them in Andromeda.
- Keep the winners. Kill the losers.

Zero creative risk. Immediate ROI.
This is the lowest-hanging fruit in AI creative, and somehow everyone skips it to go play with product insertion that looks like shit.

Don't be everyone.

What you're doing:

Time: 30 minutes for 30 animated GIFs/Videos
Cost: $60/month for Midjourney
Skill Level: If you can upload an image and click a button, you're qualified

Why this matters for Andromeda:

Your Andromeda account is probably starving for creative variations. You know the drill, you need volume to test, but production is expensive and slow.

This solves that.

You're not creating new concepts from scratch (risky). You're creating variations of concepts that already work (smart).

More tests = faster learning = better ROAS.
That's it. That's the game.

Step 1: Get into Midjourney

Go to midjourney.com. Sign up. Pay the $60/month. Don't be cheap about this, you'll waste more than that in 10 minutes of a production call.

You want the Standard plan minimum. Basic is too limited for this workflow.

Step 2: Upload Your Best Static Ads

Note: I did these live with Andrew Foxwell and Brad Ploch on their Scalability School Podcast. I highly recommend checking it out!

Start with your top 10 performing statics from the last 90 days. Don't upload random shit. Upload winners.

Step 3: Upload your image to GPT5 to help gather some inspo

Don’t overthink the prompt:

You’ll get some concepts back. Find one you like, or just come up with your own!

In this case, I wanted to add some animals in the background, so I told that to ChatGPT, and then gave it the following prompt:

Note: as always, it is important to include the copy that is on the ad + on your products in the prompt. In short: any text, include in the prompt!

Which gave me the prompt:

While ensuring I was getting an output in 9×16 and using version 7 of the model.

Hit enter. Wait 30 seconds. You'll get 4 variations of your static.

Some will be shit. Others will be usable.

Step 4: If you want to get edgy, use the vary feature

Under your generated images, you'll see buttons:

Here's what you actually use:

Vary (Subtle): Small changes. Use this when you like the image but want slightly different motion/energy. This is your workhorse.

Vary (Strong): Bigger changes. Use this when you want more dramatic variations. More chaos, but sometimes you find gold.

Vary (Region): Select a specific area to change. Honestly? Skip this for now. You don't need it yet.

The Strategy:

Step 5: Download Everything

Click each image. Click "Open in browser". Right-click and save.

Yes, this is manual. Yes, it's annoying. No, there's no better way yet for what we're doing. It takes 2 minutes per static.

Step 6: Upload to Meta in a New ASC / CBO / ABO / Cost Cap / Capsule / Dump / Words / Idk

Take all those variations. Upload them to Meta as new ad creative. Same copy, same targeting, just different motion/energy in the creative.

Let them run for 3-5 days. Check your results.

What you're looking for:

This data tells you what to generate more of.

By the end of Week 1, you should have:

✅ 10 proven static ads uploaded to Midjourney
✅ 3-5 variations generated per static (30-50 total new assets)
✅ All variations uploaded to Andromeda
✅ Tests running for 3-5 days
✅ Data on which motion styles perform better

That's it. That's the assignment.

Don't overthink it. Don't try to make perfect ads. Don't skip ahead to Week 2 (well you can’t really, because I haven’t written it yet, ha!).

Just do this. Generate variations of shit that already works. Feed the machine. See what wins.

Mistake #1: "I'll just generate completely new concepts"

No. Stop. Use proven statics first. Once you understand the workflow, then experiment with new concepts. But start with winners.

Mistake #2: "These don't look perfect, so they won't work"

Meta doesn't care if your ad is aesthetically perfect. It cares if it converts. Some of your "weirdest" variations might be your best performers. Test them.

Mistake #3: "I'll organize these into folders and create a system first"

You're stalling. Just download them and upload them. Organization comes later after you know what works.

Mistake #4: "I'll wait until I have more time to really focus on this"

You won't. You'll never have more time. This takes 30 minutes. Do it today or admit you're not actually going to do it.

That's it.

Share your weirdest AI output in the replies.

I want to see what chaos you generate.
(Bonus points if it's weirder than an infant with better work ethic than me.)

Next week: We fix the weirdness. We polish the winners. We make the plastic-looking people look like actual humans.

Remember, don’t skip ahead. Let’s build a foundation.

Look, I'm writing this at 9pm on a Thursday while my kid destroys a busy board and my dog terrorizes squirrels. If I can make time for this, you can spend 30 minutes in Midjourney.

Don't be the person who reads all 5 emails and implements nothing. Yeah, I’m calling you out. 🫵

Be the person who shows up to Monday's standup with 40 new ad variations already testing.

See you next week. I'll be the one recovering from watching an infant dressed as a dancing bear.

Your favorite AI influencer,
Will

See you next week.

P.S. Yes, some of your variations will look like everyone got Botox. That's next week's problem. This week is about volume.

P.P.S. If you skip ahead to product insertion without doing this first, I will find you.