To be clear. Without a lot of work, AI produced material doesn’t sound that good. Stems are full of artefacts and vocals sound synthetic. The extensive application of reverb is jarring. But as a demonstration of my song writing abilities it’s functional. Most of the material I’ve written with so far is withheld for proper treatment and real artist collaborations.
How I work
The spark
Every song begins with a spark: an idea, an experience, a feeling, a phrase or even a single word.
Genre and style
That spark decides the genre and the style. Sometimes an idea fits neatly into one of several existing projects with an established sound and tone. Sometimes it stays a stylistically isolated experiment.
Once it’s set, it doesn’t change. I don’t do genre roulette.
The draft
I never generate anything without a completed first draft.
Things change, of course, but every building block needed to convey the intended message and tell the intended story is in place before I go any further.
Finding the melody
If I have a melody in mind from the start, I use a few tricks to make the AI interpret it correctly: capitalisation to steer intonation, recorded humming, or actual chord progressions. If I don’t, I play around with the lyrics until something clicks. For songs that belong to an existing project, I use the project’s other songs as a reference.
This is melding words into melody in practice.
The demo and the rewrite
Next, I use AI to produce a demo that sounds roughly the way I intend. It’s. A. Process.
Then I refine the lyrics until the theme and tone hold together all the way through. Every word in every draft, revision and version is mine.
Where it ends up
Finally, there are two paths. Either I keep working on the AI-produced version and release it as a K Verse creation as a showcase. Or I pitch it to producers to be rearranged, produced and sung by a real artist. I can do some production work myself, but it’s not what I’m in this for. Not yet, anyway.
Know your lane.
