The Unsung Hero

Most people watch an ad and never hear the music.

That's the point. When the music is doing its job, you don't notice it. You just feel the thing it's quietly doing to you. The tension. The lift. The sense that something is about to happen, or just did.

Last month I made a watch ad. For a brand that doesn't exist, a watch you can't buy, a man who was generated by a machine. I've talked about that ad before.

But I haven't talked about the part I cared about most.

The music.

The music wasn't the last thing I added. It was the first thing I locked in. And then I built everything else to fit it.

The Hunt Before the Find

Nobody tells you this about scoring something with AI. The tool is fast. Finding the right answer is not.

I use Suno. I can generate a finished, fully arranged track in less time than it takes to make coffee. So you'd think the music would be the easy part. It wasn't. It took the longest.

I went through genre after genre. Generation after generation. A bolero, because I thought the ad wanted something slow and smoldering. It didn't. Too slow, and it just sat there. A Bond-style theme, because the ad has a man, a mystery, a certain kind of cool. That one didn't work at all. It felt borrowed. Like I was renting someone else's idea.

Salsa moderna. Cinematic Latin jazz. American swing with a cha-cha rhythm underneath it. The Son Montuno came close, but it moved too quickly and didn't build up the suspense or anticipatory energy I was looking to create.

Seven genres. Maybe more. I lost count somewhere in the middle.

And then I heard it. A Latin swing big band ballad. When I heard the piano part, I knew. It had the calm confidence the man needed - the same confidence the whole ad was supposed to carry. It did everything I wanted in the first fifteen, twenty seconds.

One Track, Two Jobs

But finding the genre was only half of it. The real problem was structural.

The ad has two halves, and they don't feel the same.

The first half is a man moving through his workday. There's a question hanging over it. A little mystery, a little suspense. You're watching him and you're not quite sure where this is going. The music there has to hold a breath.

Then he meets her. The second half happens on the Riverwalk, and the whole thing opens up. The mystery resolves into something warm. The energy shifts. You can see it on screen.

I wanted you to hear it too.

One ad. Two feelings. And I'd given myself one song to carry both.

So I did what any musician does when one take won't give you everything.

I edited.

I took roughly the first twenty seconds for the opening half, where the suspense lives - the stretch right before the song shifts into its higher-energy, resolved section. Then I jumped to a later part of the same track for the ending, where it blooms.

I spliced them at a seam I worked hard to hide. The track repeats a musical pattern, so I lined the first half up against the later instance of that same pattern - the one already heading toward a conclusion. Musically it just made sense. Same phrase, further down the road, pointed at the ending I needed.

And I didn't end on a hard stop. I let it fade. Because the ad ends but the moment shouldn't. The fade tells you the dancing keeps going after the screen goes dark.

Then I went back to the visuals and cut them to land on the music's markers. Not the other way around. The track was the framework. The visuals were edited to fit the music.

What the Tool Couldn't Do

Here's what I keep coming back to.

Suno generated that music in seconds. Anyone with an account could have generated the exact same track. The tool is not the rare thing. The tool is available to everyone now.

But Suno couldn't tell me the ad needed two movements. It couldn't hear that the first half had to feel like a held breath. It couldn't feel the exact moment that breath should release. It couldn't find the seam. It couldn't decide on the fade.

I could. Not because I'm special. Because I started piano lessons when I was six years old, and I've spent more than three decades since then behind a drum kit and in front of a band, learning in my body what music does to a room.

That's the part that didn't come from a tool. That came from the work. Over thirty-five years of it.

The machine handed me the notes. Knowing what the notes needed to do - that was mine.

Key Takeaways

1. Lock the emotional anchor first. Before you choose visuals, fonts, or pacing, decide how the thing should make people feel. Build the rest to serve that.

2. The tool gives you raw material, not judgment. AI can generate a hundred options in a minute. Knowing which one is right is still your job, and it always will be.

3. One source can do two jobs if you know where to cut. Don't assume you need more material. Sometimes you need a better understanding of the material you have.

4. Hide the seam, then trust it. The craft isn't in having perfect pieces. It's in joining imperfect ones so well that nobody looks for the join.

5. Your years are not a sunk cost. They're the moat. Whatever you've spent a long time getting good at is exactly the thing AI can't hand to someone else. Use it.

Final Note

The watch in that ad isn't real. The brand isn't real. The man isn't real.

But the three and a half decades I spent learning what a song is supposed to do - that's as real as anything I own. And it turned out to be the most important tool in the room.

Funny how that works. You spend your whole life thinking the skill is the thing you'll use someday. And then someday shows up wearing a watch that doesn't exist.

Until next time, 

Elliot