I don’t really care much for AI generated music tools as they are because the generated stereo digital audio is not very flexible so doesn’t provide a great starting point for music producers.
The whole industry has missed an opportunity IMO. There are weaker implementations from InstaComposer (W.A. Productions) and the Capatin plugins but I think Scaler has more potential. The usage of AI should augment the existing performances library and not interfere with it too much but AI generated performances along with the ability for users to edit them, should be a way to add them to your library.
I do agree with your comments about the current state of AI in composing, but I do think i tmay be useful to help generate ideas.
Having said this using arpeggiators with randomisation such as Stochas or ChordPotion can produce some interesting ideas. An example of this is in the video about using ChordPotion to create ARPS for Trance music. Even if Trance is not your thing his illustrates how you can use a tool such as ChordPotion to generate melodic motifs.
The beauty of tools such as ChordPotion is that you are not limited to generating simple arpeggios.
Here is a short motif that I created using Scaler and ChordPotion in less than five minutes CPEX.mid (224 Bytes)
I think mostly when people think about AI generated things they think about prompting. Of course it makes more sense when asking for something in a longer format. The existing performances library has very short loops compared to what AI would be good for using the text prompt generation. A good comparison would be EZ keys which is also library-based for the most part.
An example prompt might be “give me a 64 bar plucked ostinato with a call/answer motif that has a variation ever 4 bars”. The output gets added to the user-preset library of performances. The chord potion tool looks great but it’s not the same as using a text prompt.