124 karma · joined April 6, 2017
This is so good and yet so painful.
Also, I wonder if there will be a natural “backlash” against generic AI music and towards stuff that has more human imperfection. But then again even the AI will learn to imitate the imperfections so what exactly will make a piece more human in the long run?
Answer: 1
EDIT: I wrote this without carefully reading the article (oops!). Author does a great job with the "adjustments"...I suspect adjusting to relative minor covers a huge amount of these issues, and the author throws in mixolydian for good measure! There will always be edge cases, secondary dominants, modulations, blah blah blah, but I suspect adjusting to include relative minors handles the vast majority of popular music.
https://calvinflegal.bandcamp.com/track/when-you-were-around.
I more recently put a song on Spotify. I'm somewhat happy with it, but the vocal lacks confidence vs the first one imo:
https://open.spotify.com/track/6XlWm24ERDdgmK4Z6Px5TD?si=vkN...
Having critiqued myself as above, my next one will have a more confident vocal. Here's a tiny clip: