How I make music with AI

In the first episode of this series (What if you played at being an artist with AI?), I admitted I have albums on Spotify… even though I can’t sing Happy Birthday in tune.

Here, I show you how I actually do it, step by step.

No magic. Just a small ritual I repeat.

🎧 Step 1: understand a style I like

I often start from a track or a band I like.

I ask ChatGPT to analyze the style. A typical request looks like this:

Explain the style of the following pieces of music — what makes them catchy?

I want to understand the rhythm, the mood, what sticks in the ear.

Alternatives to ChatGPT: Claude and Grok do the same job very well.

✍️ Step 2: write the lyrics with AI

Next, ChatGPT helps me write the lyrics.

A real example of a request:

I want to bring out the hidden life of an alpine ski coach, their work, and so on, in the form of a blues.

I edit. I cut what sounds wrong. I ask for another version.

It’s a conversation, not a magic button.

🎧 Listen to the result on Spotify

📝 Step 3: ask for the Suno prompts

When I like the lyrics, I ask the tool to write the prompts for Suno:

• the style (instruments, mood, tempo);
• the lyrics ready to paste.

That way I don’t have to rewrite everything myself.

🎹 Step 4: generate in Suno

In Suno (I have the basic paid plan), I paste the prompts.

If I’ve already saved a voice I like, I can attach it to my request.

Now you can hum or clap to help Suno understand your idea. It can replace your example with an instrument. You play an instrument: bring your real bit of the piece of music and Suno will do the rest.

Suno gives me two tracks.

I listen. I keep the one that interests me most. I re-prompt.

Usually, after about four versions, I have what I want.

Alternatives to Suno: Udio and ElevenLabs Music.

📢 Step 5: share (without pretending to be a star)

For Spotify, Apple Music, and other stores, you need a distributor. Mine was Ditto Music.

But the stores don’t love AI-generated music much.

And my goal isn’t to get rich.

So today I mostly publish on YouTube, with a still image generated by AI to represent each track.

Here’s a piece I made to work through the emotion of losing a friend:

🎬 Listen on YouTube

If you want to hear what’s already online elsewhere: 🎧 my creations on Spotify.

🎨 And images?

Music is only half the game.

For images, paintings, and photos, that’s the next episode.

🤔 Is it really art?

I’m very aware of copyright issues, and of the question “is it really art?”

My point of view is something like this: without these tools, I never would have been able to reach the artist in me. I never would have let certain emotions through any other way.

🍳 The simple version

1. I get a style I like analyzed (ChatGPT, or Claude / Grok).
2. I write the lyrics with the same tool.
3. I ask it for the Suno prompts.
4. In Suno, I generate, listen, and try again (often ~4 versions).
5. I share on YouTube with a still image — without fighting music stores.