The REAL question
Marian and I were talking about AI music the other day. I resisted trying it for a while, though I was using MidJourney for text-to-image generation and its creations were astonishing. I could put music I wrote, played and produced into these still images, and later with videos using Neural Frames, but I was reluctant to hand these musical roles over to AI. Then I began to hear about Suno from well-respected songwriters, publishers and engineers – people I trusted the most for quality and creativity. Were they converts? Put aside the big glaring rights-fight about the harvested parts – they are used to “produce” the mind-blowing tracks Suno puts together with in-depth prompting from the user. I’m sure my guitar has been used thousands of times. In fact, once my guitar has been recorded these days it becomes digital data that can be manipulated to the degree that it sounds nothing like my guitar. The sound is not human, though the real spirit and passion of the performance was/is human. But when you have an ingenious program that analyses prompts, audio and lyrics and then assembles songs and productions that rival commercial radio… well… listen to these two tracks – same song – “If I didn’t have you” – but the first is a guitar vocal (stripped down) and the second is a Suno version.
“If I Didn’t Have You” – work track Listen to this first!
“If I Didn’t Have You” – Suno version Then listen to this!
I prompted Suno to create a jazz ensemble with a male singer like Michael Buble (it won’t allow named artists prompting but will find a suitable sonic alternative). In all honesty, the production it came up with is very much what I would have aimed for as producer. And… it convinced me to try the same with other songs of mine that I felt were really good but needed a better production. You can listen to them on my Entrepy page.
Marian has always like the sound of squeaks and scraps that real guitars make, and natural un-tuned vocals, non-quantized drums and synths, etc. You know…humans.