I've been doing this. I've been a poet/songwriter for a while, but I'm no musician. This lowers the bar and provides a great deal of relief from the "creative boilerplate" necessary when booting up a song from zero using a DAW, especially for me, a non-musician.
So, I get a good song by throwing spaghetti at the wall until something sticks. Then I can export the stems to the DAW and replace the AI vocals with my own. A little audio processing and mixing later and the whole song is mine.
That’s a good idea actually! The AI did a fantastic job of making a tune that sounded good and matched the style I was going for, then a lot of the iteration was changing the verses because I discovered there’s a lot of lyrics that sound extremely “cringe” when you listen back. Like nails on a chalkboard, a “you know it when you hear it” situation.
My wife wrote a song for a story she’s been working on, and honestly her sense of verse and timing gave an output with me writing a simple style prompt that sounded absolutely fantastic. But she’d spent hours writing and refining that song as it’s important to the story.
Replacing the AI vocals with my own would likely work, although there’s a certain note in the chorus that’s beyond my vocal range. I bet if I practiced it and recorded myself singing the chorus 50 times I could get one result that sounded right, though. Thanks for sharing.
So, I get a good song by throwing spaghetti at the wall until something sticks. Then I can export the stems to the DAW and replace the AI vocals with my own. A little audio processing and mixing later and the whole song is mine.