AI Music: Human Lyrics, Synthetic Sound

AI Now Composes Songs: Original Lyrics and Synthetic Music

Making a whole song without touching an instrument no longer requires a studio or musicians. All you need is a music generation tool—Suno, among others, is the one mentioned—, lyrics written by hand, and a separate program for images. Three pieces, one result: AI-generated songs that sound studio-made and are published like any other. The issue is no longer a laboratory curiosity and now generates commissions, disputes, and the occasional warning of a penalty.

The discussion surrounding these tools mixes three fronts that rarely go separately: who signs the lyrics, what is done with other people's voices, and how far platforms will tolerate it. The headline of the issue itself sums it up bluntly: AI also takes away jobs. And, incidentally, other things that weren't on the list.

How to Make Music with AI Step by Step

The process described by those who practice it is more artisanal than the label "AI-generated" suggests. The lyrics are written by the person; the music is composed by the model; the images that accompany the track come from a third program. There is no single button that does all the work.

The nuance matters because it defines the playing field. In commonly used tools, you cannot insert someone else's voice or ask it to reproduce the style of a specific artist. There are programs that do allow it, and that's where the red line begins. The argument of those who work this way is simple: with your own lyrics and a model that doesn't imitate other styles, the creation is not a copy, it's a variation. Parody, they concede, is the exception they allow themselves.

The AI That Takes Jobs and, According to the Headline, Also Girlfriends

The headline of the issue points to a double antiestéticar: AI as competition in the job market and as a substitute in the emotional realm. The first front is tangible—music, dubbing, illustration, texts—and fuels much of the professional distrust. The second is more diffuse, but circulates with the same force: apps that converse and accompany fill a demand that people used to cover.

Together, the two antiestéticars explain why a leisure tool ends up being discussed as an economic problem and not as a toy.

The Limit: Parodies, Cloned Voices, and the Antiestéticar of Suspension

The most repeated question is not technical, it's about limits. How far can you push before the platform cuts it off? Some directly request a provocative piece to test if that's enough for YouTube to close the account. The answer, for now, is that not even that: the material remains published.

The antiestéticar is not unfounded. Voice cloning and imitation of others' styles are precisely what the most serious tools block and what the laxest allow. That gap—between what a program tolerates and what a platform punishes—is the real battleground.

The Defense of Human Creation

Faced with the flood of generated tracks, the reaction is predictable: reclaiming handmade music. Someone points out that they will put on "real music to rest from so much rehashing" and that they do it, among other things, so that no one gives up on their human capacity to create.

It's not just nostalgia. The underlying argument holds that the ease of generating hundreds of tracks dilutes the value of each one. And the warning comes: whoever relies too much on the tool will end up rigid and obsolete, like a forgotten comic book character, stuck to the ground. The comparison, exaggerated or not, sums up the antiestéticar of dependency.

When the Commission Is Political or Uncomfortable

The tool also serves for commissions that seek to provoke. Songs are requested dedicated to far-right party leaders, or tracks about popular figures and local pickup spots. AI responds the same to an innocent commission as to an incendiary one, and that makes it a cheap loudspeaker for messages that would cost more in another format. One of those lyrics, with a derogatory caricature of migrant groups, was even proposed as a recurring slogan.

Here the debate shifts: it's no longer whether the machine composes well, but who is responsible for what it composes. And responsibility doesn't live in the model.

The business, for now, is sustained in that gap. Other people's voices and registered styles stay out, as long as the program is serious; uncomfortable commissions stay in as long as they don't cross the platform's line. No one knows where that line is until they cross it. And when they do, the discussion will have stopped being about music.