FORM:

The form of AI made songs are typically very flat. The curves and changes in emotion and intensity that you would normally find in Human made music appears far less in AI. These songs typically lack a complex intro, and instead, they tend to just begin. The same goes for the outros too, even sometimes they lack them entirely. Chord progressions can also sound unnatural and also lack resolution.

INSTRUMENTATION:

Because the songs are generated from scratch entirely, the instruments are not 100% realistic. Drums are the easiest to spot due to their jittery/fuzzy sound. Cyballs are hard for the AI to perfect due to their volume and complex range of frequencies, and this goes for things like guitar and keyboard as well. Keyboard sounds may have a slower “atack” of a note then what a real piano would sound like. Guitars can also have a hard time forming complex chords and melodies, which usually results in a basic sounding track.

VOCALS:

Because of the dynamic ability of human singing, AI struggles especially with vocals. They can turn out fuzzy and muffled, and also unnaturally on pitch. Accents and words with sounds like “sh”, “th”, and “tion” can come out undefined and almost blurry. When asked to imitate an accent, AI ends up mistaking and mixing several others.

Short (1 min):