AI music is getting good. Is that good?

Describe a genre, a mood and an instrument, and a minute later you have a complete piece with reasonable arrangement and sound structure.

Technical quality is no longer the question. The question got harder: what does it mean for music production to become free?

Where it genuinely works#

There is an entire market that was suffering and has found its answer.

Background music. Corporate video, short advertising, podcasts. Here music is a function rather than an art, and the job is to fill space without stealing attention.

Prototyping. A director who wants to hear the feeling of a scene before commissioning a composer. Excellent use, and it competes with nobody.

Adaptation. The same piece at three lengths for three platforms.

And let us be honest: this market was already served by cheap, repetitive stock libraries. Who loses here is not the artist composer. It is the stock library.

Where it does not work, and this part matters more#

Music anyone is supposed to remember.

The song that accompanies a generation is not a set of technically correct decisions. It is a specific person, at a specific moment, choosing to say something in a way it had not been said. And the model is trained, by definition, on what has been said before.

The economic paradox: abundance lowers value#

Here is the point many people skip. When producing something becomes free, its value does not rise. It collapses.

When everyone can produce decent music, basic quality becomes an entry price rather than an advantage. What distinguishes becomes something else entirely: who listens to you, and why they care.

Rights, without oversimplifying#

The legal position is unsettled globally, and I will say what is known confidently and what is not.

Known: models were trained on protected works, and this is the subject of active litigation in several jurisdictions.

Unsettled: whether training constitutes fair use, who owns the output, and whether copyright protection applies to work with no human author.

Practical: if you use generated music commercially, read the platform's own terms carefully. Some grant you clear commercial rights and some retain rights that will surprise you, and the difference is not in the price but in the sixth line of the terms.

What this means for you#

If you make content: use it for backgrounds without hesitation or guilt. It is an entirely legitimate use that saves you budget and time.

If you are a musician: the exposed part of your work is the repetitive functional part. The protected part is your own voice and your relationship with the people who listen. Invest in the second.

If you teach music: a student who can generate a piece in a minute needs a question from you that was not urgent before: why this note and not that one? Justification now matters more than execution.

In closing#

Generated music solves a real problem for anyone who needs sound to fill a space, and solves nothing for anyone who wants to say something.

And when every sound is available to everyone, the only question left is the oldest question in art: what do you want to say?

Common questions#

Is generated music good enough for commercial use?

For backgrounds, corporate video and podcasts, entirely. For work an audience is meant to remember, no, because the model produces the average of what has been heard and the average is not memorable.

Who actually loses from this technology?

Cheap stock music libraries more than artist composers. The market served by repetitive library tracks is the first thing replaced.

What is the copyright position?

Unsettled globally. It is known that models were trained on protected works and that this is under active litigation. Practically, read the terms of the platform you use, because commercial rights differ substantially between them.

Why does music lose value as it becomes easier to produce?

Because value flees the layer that became free toward what cannot be copied: taste, selection, and relationship with an audience. The ability to produce turns from an advantage into an entry price.

What should a music student learn now?

Justification before execution. A student who can generate a piece in a minute needs to answer why this note rather than that one, a question that was less urgent when execution itself was the challenge.