Issue 13 · Nov 2–8, 2026Editions · Glossary · Archive · Method · RSS · HU / ENG

News of the AI era, curated weekly

Glossary › Generative media

music generation

basicspreading

Music generation is when an artificial intelligence creates new music, including vocals and lyrics, on its own based on a text description or other guidance.

In plain terms

Imagine you have a band that has listened to every possible style and can play anything you ask. You type: 'cheerful summer pop song about Lake Balaton, female vocals,' and within a minute or two the recording is ready. Music-generating AI works roughly like this.

You can quickly and cheaply create music for ads, videos, podcasts, or games with it, and it is also good for testing song ideas and personalized greeting songs. You do not need to read sheet music or play an instrument because you formulate the request in plain language.

A streaming service's recommendation system selects from existing songs, while music generation produces a recording that did not exist before. The system does not splice together song fragments: it learned from vast amounts of music how a song is typically structured and creates new material from this knowledge.

How it works

  1. Training: the model processes enormous amounts of music, often together with associated descriptions (style, mood, instruments), and learns the patterns of rhythm, melody, harmony, and sound.
  2. Request: you specify in text what you want, for example genre, mood, tempo, instruments, or even lyrics. Some tools also allow you to provide a hummed melody or a sample recording.
  3. Generation: the model assembles the sound step by step by breaking it into small units, or gradually 'cleans' the music starting from noise, while continuously adapting to your request.
  4. Audio rendering: the internal, compressed representation is turned into a playable audio file; many tools include sung vocals as well.
  5. Refinement: if you are not satisfied, you can request a new version, extend the song, replace a part, or split it into separate tracks to continue in a music editing program.

An everyday example

You are looking for background music for a video of your family holiday, but the video-sharing platform would block your favorite hit for legal reasons. In a music-generation app you type "laid-back, acoustic guitar summer tune, 60 seconds" and use the resulting music under the video.

Who is affected

marketerYou can create custom background music for ads or social media videos in minutes, but first check whether the tool allows commercial use. → Marketing
employeeIf you work as a musician, composer or sound engineer, you gain a direct competitor in the market for commissioned background and advertising music, while you can also use it yourself for brainstorming and demo creation. → Work & HR
parent, student, teacherYou can use it to create a learning song, music for a school project, or a music theory illustration, and it offers a good opportunity to discuss what creation and authorship mean. → Education & Family
everyoneOn streaming platforms and social media, a growing number of songs are made with AI, and often nothing indicates that they were not written or sung by a human.

A common misconception

MythMusic created by machines is automatically royalty-free, so anyone can use it freely for anything.
In realityUsage rights are defined by the terms of the respective service, and these often allow commercial use only with a subscription. Moreover, copyright in most countries, including the United States and the European Union, requires human creative contribution. Music created purely by machines is therefore typically not protected, meaning you may not be able to take action against others copying it.

Limits and risks

Similar, but not the same

Related concepts

Broader

generative AI, audio generation, generative media

Narrower

text-to-music, singing voice synthesis, lyrics generation

On the edge

image generation, video generation, speech synthesis, voice cloning, diffusion model

A brief history

One of the first computer-composed works was the 1957 Illiac Suite, which produced sheet music rather than sound. Among neural models generating finished audio recordings, OpenAI Jukebox appeared in 2020, followed in 2023 by Google's MusicLM and Meta's MusicGen, text-to-music systems. The technology truly reached the general public in 2024, when applications like Suno and Udio enabled anyone to create full songs with vocals.

Where things stand

Apps that turn text into complete songs with vocals are available to anyone, and machine-made music has appeared in large volumes on streaming services. Some of the copyright disputes ended in licensing agreements with the major labels, but the legal situation is still open on many questions.

Status: 2026 November

In the news

Appears in 14 Agent Age stories; the curve shows weekly occurrence in the news magazine.

Also known as

music generation · MI-zene · AI-zene · AI-generated music · generative music · music generation · AI music · text-to-music

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.