My Shadow Quit began with a very specific assignment I gave myself: create a song for the Luxembourg Song Contest 2027. I wanted something original, memorable and slightly absurd, but with enough musical substance to stand on its own. We brainstormed around fifteen unusual concepts before I chose this one. It appealed to my Magritte side: take one impossible proposition, accept it without question, and see what happens.
My first important decision was about perspective. I wanted a female narrator, a shadow with a mind of her own, and absolutely no explanation of the impossible premise. I also wanted a final twist that would arrive almost casually. The ending, with its understated “Oh. Right.”, became essential to the song’s identity.
The lyrics went through several revisions with ChatGPT. Early versions were too long, and I was particularly dissatisfied with the second half of the second chorus and the final chorus. I kept the opening lines I liked, rejected the weaker passages and worked towards something shorter, more conversational and less explanatory. By versions 9 and 10, I was satisfied with the words. The challenge was then to make the music deliver them properly.
I chose a soulful swinging groove around 116 BPM, with a warm, low-register female voice. Mabel became my preferred Suno voice for the song. I wanted confident, slightly dry delivery rather than theatrical comedy. The arrangement centred on tight live drums, melodic bass, clean rhythm guitar, warm electric piano and restrained percussion, with just a little brass colour.
Getting the balance right was surprisingly difficult. Some generations sounded too much like a fairground, with overly enthusiastic brass. Others introduced backing vocals answering the lead singer, something I specifically didn’t want. I repeatedly adjusted the production brief to keep the swing sophisticated, the humour understated and the backing vocals limited to subtle harmonies.
Duration was another constraint. The Eurovision submission could not exceed three minutes. Version 5 produced recordings of 3:00 and 2:59, but the approximately seventeen-second introduction was excessive. Later versions reduced that to around fifteen seconds. I also wanted fast transitions, without unnecessary instrumental breaks or solos.
At one point I decided the arrangement needed an unexpected instrumental signature. I considered bass clarinet, but rejected it because the sound was too low and rather flatulent for this particular song. Pizzicato cello offered a better possibility: dry, woody, rhythmically playful and just unusual enough. I incorporated it into the later production brief without allowing the arrangement to become orchestral or comical.
There were at least twelve numbered Suno versions in the documented development. Version 8 was an early favourite, while versions 9 and 10 settled the lyrics. The later experiments concentrated on refining the instrumentation and overall feel. I was looking for a recording that sounded effortlessly playful rather than one that worked too hard to be funny.
The song was eventually submitted to Luxembourg with both vocal and instrumental recordings. Preparing the application also introduced me to a useful new production technique: extracting separate vocal and instrumental stems in Suno, something I subsequently adopted as part of my regular workflow.
More than 800 songs competed for eight places, and mine wasn’t selected. That was disappointing, but the experiment had already served its purpose. I had pushed an unusual idea through repeated lyrical, musical and technical decisions, and learned something practical along the way.
The most important discovery was that an absurd song works best when I resist the temptation to make the music sound absurd.