I'm an artist with a catalog of published instrumental tracks that, for years, I've wanted to complete with vocals—a part of the production process I always found challenging. I decided to use Suno AI's Remix feature not to create from scratch, but to collaborate with the AI as a vocalist, using my own high-quality productions as the source material. This project is a case study of that workflow, showcasing the "before" instrumentals and the "after" vocal versions.
The most surprising discovery was the psychological impact. Hearing a polished, finished version of an old idea almost instantly is a powerful creative catalyst that has me digging through my entire back catalog. To achieve these results, I developed a system I call "Suno - assist," where I use another LLM to generate detailed style prompts and lyrics with embedded musical directions (like "[full band enters]" or "[ethereal falsetto]"). The page shares this exact workflow and the system prompt for anyone to use.
Beyond just the music, this experiment has me thinking about the future. The speed of iteration is a superpower for any working composer, but it also points toward a future where we may need a "Proof-of-Human" ledger to verify unassisted art. I'm excited by the possibilities and would love to hear HN's thoughts on where this is all heading.
I'm an artist with a catalog of published instrumental tracks that, for years, I've wanted to complete with vocals—a part of the production process I always found challenging. I decided to use Suno AI's Remix feature not to create from scratch, but to collaborate with the AI as a vocalist, using my own high-quality productions as the source material. This project is a case study of that workflow, showcasing the "before" instrumentals and the "after" vocal versions.
The most surprising discovery was the psychological impact. Hearing a polished, finished version of an old idea almost instantly is a powerful creative catalyst that has me digging through my entire back catalog. To achieve these results, I developed a system I call "Suno - assist," where I use another LLM to generate detailed style prompts and lyrics with embedded musical directions (like "[full band enters]" or "[ethereal falsetto]"). The page shares this exact workflow and the system prompt for anyone to use.
Beyond just the music, this experiment has me thinking about the future. The speed of iteration is a superpower for any working composer, but it also points toward a future where we may need a "Proof-of-Human" ledger to verify unassisted art. I'm excited by the possibilities and would love to hear HN's thoughts on where this is all heading.