I promised @jim earlier that I would attempt to create a song that can be played on the radio using nothing but AI tools. This has taken a lot of effort over a few months, and has been delayed by developments in the litigation I'm working on, but I have an "alpha version." The idea is here, but still has a long way to go to get the little things right.

The overall goal is to intentionally use nothing else but AI tools to generate something unique that people do not think an AI tool can generate and which would not normally be heard in pop music. Otherwise, it wouldn't break through the noise. Therefore, the song could not just be a standard 2020s major key Taylor Swift singer-songwriter song, as that basically comes out as the default from these tools.

https://shoemakervillage.org/temp/persona_falsa_alpha.flac

I want some comments to see if this is going in the right direction. I'm already aware of many things that I want to fix, most notably:

• The conclusion is obviously incomplete

• The last chorus needs more emotion and less stiltedness in the delivery. Some sections need more emotion in general

• More internal variety within the various sections

• Stereo width issues in places

• The latin words like "spectatus" are not being pronounced correctly in some instances, and the volume on them is too low at the end

• Some parts are too "perfect" - an uncanny valley because everything is exactly on key and on beat

• There needs to be slightly less vocal processing and more reverb in the very first verse

• Many parts lack good transitions and sound out of place, like the first part of the second verse - because they were indeed actually pasted together

But this is far enough along now that it's more than an "idea," so it can be evaluated. Please post your comments!