On my Sunday Night

Hi! I'm DoctorBlackstone, feel free to call me Doc.

I'm an AI Music Creator. If you are a curious person, you might enjoy this one.

I'm not here to argue whether AI music is real music, or art, or not. I have my opinion on that, you have yours, all good. Today I want to talk about something else: the idea that "AI music is easy".

Let me get one thing out of the way first. I'm not a musician, even though I play the sax. Studying an instrument for years, or living in a studio for months to finish a track, that's for real musicians. I think of myself as a hobbyist, and a bit of a scientist, given my academic background.

So, "AI music is easy".

Most people picture it like this: you type "Stairway to Heaven", hit "Generate", and post whatever comes out on SoundCloud or Spotify. Some creators do exactly that.

I want to be perfectly clear: that is NOT what I do.

And for once, instead of just telling you, I'm going to show you.

As a thank you for your curiosity, I'll let you hear something I've never shared before. The master of one of my songs. This song. The raw starting point, before I do what I do. Spoiler: "Sunday Night" was a LoFi with Harmonica.

But before that, some context.

Here's some of the data I've been collecting since December 2025.

This is Experiment #20440. That number is not a gimmick. I've genuinely generated over 20 thousand audio files with AI. I still have every one of them.

Out of those 20 thousand files, only 35 became released songs. That means 99.83% of what I generate never sees the light of day. But that doesn't mean it's wasted. Pas de Deux, for instance, was born from an idea I tried on Claire.

I call those "experiments", or "rounds". It's a trial-and-error process of ideas and edits that I run on a "master" until it sounds the way I hear it in my head. I only stop when it hits a sweet spot, or implodes completely.

That gives an average of 584 to 1. For every 584 experiments, only 1 becomes a released song. But that's on average.

Experiment #9296 (Some People): 1915 rounds. And I still hate that fucking Clown (great song though).

Experiment #14649 (Morning News): 1564 rounds.

Experiment #10873 (Tell Me): 935 rounds.

This one you're about to hear, Experiment #20440 (On my Sunday Night): 520 rounds.

Each experiment is an entire song. This means I have 520 different versions of "On my Sunday Night". My songs average 3:59, so for this track alone that's 2,071 minutes of music, or 34.52 hours. And I listen back to check that every change I made actually landed the way I wanted.

I like to lean on the side of honesty, so I should say there are cases where I don't listen all the way through, because a round got declined early. That usually happens when a file was mis-generated, or failed at something the song needed.

But broadly speaking, I've generated more than 1,300 hours of music. Almost two months of it, back to back.

All of that to say: what I do is not one-and-done.

Working with AI music feels a bit like sculpting or molding. I build from the ground up until I typically have all the layers a song needs. Yes, layers. Experiments like #6618 (Do or Die), for instance, have over 34 layers. I work each layer until it gives me what I want from it. There's a real, time-consuming process behind that whole trial-and-error.

And that's just for the songs. I won't even get into the trial-and-error of writing lyrics, how hard it is to keep a consistent visual identity for the character, the bills from the multiple AIs I use, Groover and SubmitHub, and all the rest. I'll save that for another time.

Like I promised, here's the Master. Version 1, first generation, untouched. The only thing I did to it was remaster it. Nothing else. This was my starting point for "Sunday Night" that then became "On my Sunday Night". It's right below, next to the finished track.

Listen to both, then draw your own conclusions.

To wrap it all up: none of these numbers make the songs good. Whether they are any good is for you to decide. All I ask is that you don't dismiss them just for being AI music.

And to all the partners, friends, and listeners who understand what I do, who in any way enjoy what I do, and who have been with me since last year: thank you, as always.