Tag: Recording

Excuse the meat-related analogy, but when you listen to a song on streaming services or elsewhere, it’s important to remember that you did not see how the sausage was made. You are not hearing the songwriter’s multiple poorly recorded voice notes or all the wonky demos that did not make the cut.

More often than not, this is the pristine, glamour version that has been honed and re-recorded several times, featuring comped vocals. It also benefits from the skills of an engineer using expensive plugins or outboard gear to massage the tracks and the output, as well as the work of a mastering engineer who has sprinkled the mixdown with his own brand of musical MSG.

Using this method, an average song can sound pretty good. But you are not aiming for average in the long run (although don’t be afraid to write average songs on your search for the good stuff).

The lesson is: don’t try to sound shiny at the writing stage. Concentrate purely on the quality of the arrangement, lyrics, melody, texture, feel, emotion, etc. Knock the song into shape without worrying about how it will sound on Spotify. Song first, production later.

Eventually, you will play your shiny master to someone, and they will have no knowledge of the toil that went on behind the scenes. But you will.

So far, my feeling about AI is that, beyond its usefulness as a search engine, it’s only helpful for the last one percent of most creative work.

Let’s say you’ve written your lyrics (or article, or speech etc.). All the ideas, research, content, and phrasing come from your own brain. Then, you feed it into an AI tool of your choice, and it corrects your grammar or highlights awkward logic and inconsistencies.

AI is amazing at this final stage. It can catch and polish things like only a robot can.

But in my opinion (so far), the “artificial” nature of its “intelligence” becomes very evident when trying to create art.

A little experiment to demonstrate

I once wrote a song called Utopia that ended up on my last album Healer. You can listen to it here:

I started the song while living in southern China, in a factory town called Humen. Most of the inhabitants were migrant factory workers. It was an atmospheric place with a lot of history, but the pollution from the surrounding factories was chronic, and the workers didn’t look happy to me. This inspired the feel of the song and the lyrics.

I wanted to paint a picture of people trying to— or pretending to—be happy (“this is our Utopia“) while living in a pretty soul-destroying environment.

The final lyrics took a long time to write and included references to this art exhibition: Whose Utopia?. I also mention a burning motorcycle I saw in the middle of an intersection during a morning run. The coda at the end references laughing gas canisters strewn across a park in an affluent part of Manchester, where I stayed during COVID.

Here are the full lyrics:

Utopia

Oh what a pretty skyline 

Fog and pollution 

People lining up for execution

Silent ballerina 

Dances through a factory 

Lighting up the workers and machinery 

This is our Utopia 

This is our Utopia 

Don’t shatter the illusion 

This ignorance is opium 

Oh how can the morning 

Sky look suicidal 

High above a burning motorcycle

And lonely out-of-towners 

Carrion for vultures 

Biting on the hand that never feeds us 

This is our Utopia

This is our Utopia 

Don’t shatter the illusion 

‘Cause this ignorance is opium 

Don’t shatter the illusion 

‘Cause this ignorance is opium

There’s laughing gas 

On the summer grass 

This is our Utopia 

The question is: Could AI have written these lyrics or something that I consider of similar quality?

Let’s see:

Here’s the result: 

Utopia (Chat GPT version)

Neon hum on the factory floor,
midnight shifts and metal doors.
Machines sing their static tune,
under an artificial moon.

And the faces glow in flickering light,
dreams wrapped up in cellophane tight.
Is this paradise? Is this a dream?
Or just a town caught in the machine?

Utopia—painted in smoke and steel,
we laugh like ghosts so the world won’t feel.
We chase the sun that never sets,
inside a dream we can’t forget.

Plastic gardens, concrete skies,
we build the world with tired eyes.
One more stitch, one more spark,
one more light against the dark.

And the radio plays a hollow tune,
echoes bounce in an empty room.
Is this paradise? Is this a dream?
Or just a town caught in the machine?

Utopia—painted in smoke and steel,
we laugh like ghosts so the world won’t feel.
We chase the sun that never sets,
inside a dream we can’t forget.

Outside the walls, the stars still burn,
but we’re too tired to lift our heads and turn.
So we dance in circuits, loop and rewind,
finding joy in the factory grind.

Neon hum on the factory floor,
midnight shifts and metal doors.
Maybe someday we’ll break free,
or maybe this is all we need.

Well, they aren’t exactly terrible.

But of course, it’s the melding of the lyrics with the melody and chords that makes the song.

I played with Utopia’s chords—particularly the chorus—for years before I got it right.

But let’s see what AI comes up with. This is the version created solely with AI using Suno.com. It uses the Chat GPT lyrics above:

And this is the version also with Suno.com that uses my lyrics:

I think the results speak for themselves, but that’s just my opinion.

Finally, what about album artwork?

This song was originally going to be on my previous album, Underground Man. My artist friend Johnnie created the album artwork.

The “prompt” I gave him was:

A psychedelic scene based on the Shanghai skyline, with pollution oozing into the river in unusual colours.

And this is what he sent me:

Which became this:

But what will AI come up with?

This time let’s ask Copilot for help using the same prompt:

It’s pretty cool. But very AI.

So what do you think? Should I have used AI to create my music and artwork?

My most-streamed composition so far was written in the Nicaraguan mountains. In 2018, I lived in a town called Diriamba. We stayed on the second floor of a large house in town. The weather was sultry, and the noise of funeral processions, bar fights, farmers on horseback, and the crowing of a giant mechanical cockerel—an evil mascot for a chicken restaurant, blaring even at 5 a.m.—poured into the room.

I wrote two pieces there: Olvidar, which started on electric guitar, and a long ambient piece designed to help my wife sleep, which eventually became Silent Vapour Trail. Both tracks appear on one of my “travel” albums, Connections.

Now, Silent Vapour Trail has only had a modest number of listens across platforms, but I once uploaded it to SoundCloud under the name Relaxing Music to Help You Fall Asleep. It’s now had over 4,000 listens—which is a lot for the non–Justin Biebers of the world, like myself.

That must be down to the name—it promises something useful. But the thing is, the music does help my wife and me sleep. Life was stressful before we arrived in Nicaragua, so I deliberately wrote a long, meditative piece to help us (and anyone else) chill out.

Still, it makes me wonder—what if all my songs had titles that promised something? Love Song to Make You Feel Romantic, Upbeat Tune to Get You on the Dance Floor, Thoughtful Folk Song…

Maybe that would help listeners who don’t know or care about my music. But it wouldn’t feel honest. I called it Silent Vapour Trail because it sounded poetic, it suited my mood at the time, and I could see airplanes drifting across the pink Nicaraguan evening sky as I wrote it.

Then again, maybe none of that matters. Somewhere out there, a tired soul has drifted off to sleep listening to whatever this piece is called, blissfully unaware of the song’s origin—or its writer.

Sweet dreams.