Dead Channel started as an idea and then got slightly out of hand.
What began as creating a fictional band gradually became albums, solo records, acoustic sessions, artwork, sleeve notes, old demos, side projects and an increasingly detailed history of five musicians who don’t actually exist. Somewhere along the way, Rika, Kai, Yumi, Mika and Ren stopped feeling like a collection of character descriptions and started feeling like a band whose history I was uncovering as much as inventing.
That’s probably been my favourite part of the project.
The songs weren’t created by simply asking AI to “make a rock song.” I spent a lot of time working out who these people were, what each of them would write about, how their tastes differed and how those tastes might change as they got older. We developed distinct voices and musical identities for the individual members, along with recurring instruments, relationships, influences and little bits of personal history. A song for Mika shouldn’t sound like something Rika would write. Kai approaches a guitar differently from Rika. Dead Channel’s first album shouldn’t sound like the record they make years later after they’ve lived separate lives.
From there, the process became surprisingly similar to how I imagine producing a fictional record might feel. I’d start with an idea for a song—sometimes just a title, situation, riff concept or something happening in one of the characters’ lives. ChatGPT helped me develop the lyrics, arrangements and production direction, and then I’d take those ideas into Suno and start experimenting.
And there was a lot of experimenting.
Songs would get regenerated because the voice wasn’t right, the guitar was too polished, the tempo felt wrong, a chorus wasn’t landing or the whole thing simply didn’t sound like Dead Channel. Sometimes an unexpected generation would suggest a completely different direction. Other times I’d have something that was almost right and spend ages trying to coax the final version out of it.
Eventually I realised I wasn’t really approaching the project as somebody asking an AI to generate songs. I was approaching it like I was producing the albums.
I was deciding which take worked, which songs belonged together, what should open an album, what needed cutting, where the record needed something faster or quieter, which version became the album version and which odd alternative take got thrown onto the fictional rarities pile. Track listings changed. Songs migrated between albums. Covers and back covers were made. Sleeve notes started explaining why records sounded the way they did. There are demos, live versions and acoustic arrangements of songs that exist because it made sense for the band to have made them.
That’s an absurd amount of worldbuilding for a fictional rock group.
It’s also been enormous fun.
One of the strangest pleasures has been gradually learning about the band while making it. There are things about these characters that weren’t planned at the beginning. Their personalities emerged through the songs and stories. Mika’s eccentric little repair shop, Rika’s battered old Heartbreaker guitar, the complicated but affectionate history between Rika and Kai, Yumi’s life away from the stage, Ren quietly being the musician who helped turn a bunch of teenagers into something resembling a proper band—all of that accumulated naturally.
The later records became more interesting because the earlier ones existed. A lyric could refer back to something that happened years before. An acoustic arrangement could reveal something different in an old punk song. A terrible teenage demo could sit beside the accomplished musician that teenager eventually became.
None of this is pretending that Dead Channel are a real band. They’re fictional characters in an ongoing creative project, and the music and artwork have been made using generative AI tools as a fundamental part of that process.
And I completely understand why that means some people won’t be interested.
AI-generated music and artwork are contentious, and there are perfectly reasonable discussions to be had about what these technologies mean for musicians, artists and other creative people. I’m not trying to convince anybody that they should like AI art, nor am I presenting myself as a musician or illustrator because I made this project.
In fact, almost the opposite is true.
I’m not particularly musical and I’m not an artist. I couldn’t personally perform these songs, produce these recordings or illustrate these album covers. Before tools like these existed, an idea on this scale would probably have remained mostly inside my head. Being able to take that idea and actually explore it—to hear approximately what these fictional records might sound like, see the band, design their albums and follow their story across years of imaginary music—has been a genuinely enjoyable creative experience.
The technology didn’t remove the part I enjoy. It gave me a way of doing it.
I’ve still spent far too long deciding whether a bass line sounds like Yumi, rejecting versions because Rika suddenly doesn’t sound like Rika, rearranging imaginary album tracklists, arguing with myself over fictional recording chronology and creating backstory for why a song recorded by people who don’t exist has three different versions.
Apparently that’s my idea of fun.
If you enjoy Dead Channel, brilliant. Hopefully exploring their music and history is at least a fraction as entertaining as making it has been.
And if the use of AI means this simply isn’t for you, that’s fine too.
No hard feelings.
There are plenty of other channels.