Lately I’ve seen a lot of people push back against AI touching their creative work. I get where they’re coming from, but I’m in the camp that likes to experiment. Take my own work as input, add some randomness, and suddenly you’re pulling surprise after surprise out of a blind box. So much fun!

It reminds me of a famous Nagoya dish called “hitsumabushi” (grilled eel rice, three ways). They slice grilled eel into small pieces and claim you can enjoy three different layers of flavor by eating it in a specific order with different toppings. First plain (round one), then with wasabi and green onions (round two), then with dashi broth poured over it like ochazuke (round three). I think it’s mostly a gimmick, though having something to do while eating does keep people off their phones, which probably does make it taste better.

But with AI, I realized my old music can get the same treatment, times N!

I previously shared the music I made in ten days years ago. I also mentioned that as a beginner, my mixing and mastering were pretty bad.

But AI doesn’t care about any of that. It just needs the concept.

A few months ago, when I learned Suno could accept input files, I fed it to Suno v5. I ended up playing with it all day.

Here’s my one song, seven ways!

Round 1! Original

Just… this. Honestly, I’ve listened to it enough times that I’m a bit tired of it. I dug up an old journal entry where I wrote this about it at the time:

I think I’ve far exceeded my own expectations, but I’m still not very satisfied with the finished product. I prefer the kind of music Spangle call Lilli line makes: simple yet layered. This song, I just kept stacking tracks on top of each other. Too many melodies going at once. My brain got a nice massage, but it’s still too messy. And I don’t understand music theory at all, so the melodies don’t complement each other. I still have a lot to learn. — April 10, 2020

Round 2! Piano solo

It kind of reminds me of Yukie Nishimura, whom I used to love. Just a little bit, but it’s so soothing.

Round 3! Rock

I originally wanted to make a math rock track. So hearing how AI made it even more rock was super interesting.

Round 4! Christian worship

The default style when I gave it zero prompting. It sounds like the kind of songs I used to sing along to when I showed up at church in college for the free food. Church songs are always so good. This one fits right in. The lyrics are all about “praying” and needing God’s guidance so you won’t be alone. The only thing missing is a “My mighty savior” thrown in. The fact that my original “Men of Determination” main melody could turn into this is pretty incredible.

Round 5! Chinese rap

My prompt was going for something like Jay Chou’s “Wounds of War.” The lyrics landed in a similar ballpark, also anti-war, but the delivery came out way more aggressive. Not bad, though.

Round 6! J-pop

I actually prompted for something YOASOBI-ish, but it came out sounding more like LiSA. Could totally work as an anime opening. The lyrics are classic J-pop: vague evocative words like “starry sky,” “heartbeat,” “dreams,” plus every variation of “kimi” — “always thinking of you,” “I want to reach you,” “don’t give up.” Perfect.

Round 7! K-indie

This one blew me away the most. I prompted it in the style of SE SO NEON. Unlike the others, if this came up on Spotify Discovery I’d actually hit Like. That guitar solo at the end is incredible. The lyrics were just whatever ChatGPT came up with, so I have no idea if actual Korean speakers would find them hilarious.


I don’t know what other people think when they hear these, but as the original creator, getting to hear my own work covered in so many different styles is genuinely one of the happiest things.

This used to be a luxury reserved for only the most popular songs, like TK from Ling tosite sigure’s “unravel” or LiSA’s “Gurenge”. The fact that a nobody musician like me gets to experience this too? I can only say it’s incredibly lucky and incredibly fun.

Every time, I’m also curious where AI decides to add new sections and transitions, which parts it keeps, and which parts it cuts. These choices always bring me new inspiration.

Hearing my own messy melodies put to good use feels amazing.

The slightly sad part is that the more it gets remixed, the more I like it? So the main issue is really that the original has too many problems, just like I wrote in that old journal entry: I still have a lot to learn.

And that’s the experience of covering my own music with AI, one song, seven ways.