AI Music Generation

I’m just catching up on some of the hyped AI apps from the last week or so. One that’s got a lot of attention is music generator Udio. I’ve had a quick play with it, and it’s fun and easy to use. I think the results are perfectly serviceable for some use cases.

Basic music generation is one of the easier AI tasks – it’s history goes back to the 1950s, and Brian Eno coined the term ‘Generative Music’ in the 1970s. It’s because you can generate ‘listenable’ music by following a small number of rules.

Full disclosure here – outside of work I’ve always written and played music, and I use a lot of technology and always have, going back to sequencers running on Atari STs. A mix of guitar and keyboards with a mix of both analogue tools (include a vintage analogue synth and old effects pedals) and digital (modelled effects, software synths, and everything modern recording software can offer).

As with most genAI, I think it’s really about the right use case. At the moment, genAI music is unlikely to change listener’s lives in the way the best human art does, but its capable of generating perfectly acceptable background music for, say, a corporate video. Obviously this isn’t great news for musicians that earn money licensing money for such uses.

It’s also a lot of fun, and opens up music creation to people who haven’t got round to learning instruments, some music theory etc etc. This, in my view, is a great thing. Lowering barrier to entry has often led to explosions of creativity when people take the tech into new and unexpected directions (synths, sampling etc), or even just get the bug after seeing that they too can do it (hello everyone who might have been at the Pistol’s Manchester Lesser Free Trade Hall), and I’m pretty sure we’ll see the same with AI music.

Interesting to see, Bandlab (one of my favourite music tools) already offers an AI generator to help musicians get over the ‘blank page’.


So dive in, play, and maybe you’ll get the music making bug too.

https://www.udio.com/