I have several different thoughts on AI music.

AI Music tools

Most of the online tools are horrible, including from well-known AI companies. There are a few that are OK or produce good enough music:

• Suno.com good overall but horrible response rates, even for the billing email address. Clearly these folks don’t care about customer service in the least. Also, got an incomprehensible response regarding copyright infringement claim, based on public domain content.
• aimusic.so, about as good as Suno though can’t produce the full range of vocals and vocal styles that Suno can. Free account cannot download songs, have to record through audio card (simple with Audacity).
• Moises.com, haven’t played with this much, but they seem to have trained their voices on copyrighted material with permission.

There are quite a few other tools out there, but many are crap. The market for generative AI in music is huge and growing.

• Grandview Research - Generative AI in Music report
• Musicful.ai - AI music statistics

The bad

• Trained based on copyright infringement.
• Use might itself be copyright infringement (derivative works).
• It is hard to get clean IP out of anything touched by AI, though enough human involvement should work for copyright (USPTO).
• People hate it and won’t listen to it (instead of giving a listen first, they simply ignore it).
• Cannot actually perform the music once it is created by AI (or have to reverse-engineer the music to be able to re-create it).

The good

• Can make banging songs, with one’s own lyrics (Odds and Ends).
• Can revitalize public domain poems into banging songs (Poems for Poets, Rabindranath Tagore).

The DAW problem

• Most Music AI tools are either one or more tools to do various things (such as stem separation) or wants to be your DAW (and moreover an online / cloud DAW).
• To deal with this, it is best to move everything into one’s own tools, and deal with AI as simple plugins that can help with the creation process, but not replace actual human interaction with musical instruments and devices.

In my case, tools of choice are Audacity and Reaper, as well as OBS studio and MuseScore.

• One example of AI for Reaper is Reaper Reapy MCP which allows for chat / prompt control of Reaper (which has a lovely Python interface).
• Another example is an AI Equalizer plugin (as usual the term AI is stretched and crumpled to mean, basically, anything).

The lyrics problem

• Personally, this is not a problem I have, as a poet and lifelong reader and writer. However, it appears to be popular. Of course we can’t expect much insight from LLMs which is what are used for this part of generative AI in music.

Generative AI for ideas

• Possibly the best approach (and one likely being used by the majority of musicians) is generating ideas on what to do with a work-in-progress.