We’ve been through LLMs, Image Generators, looked at AI Video and explored AI voices, it’s time to see what generative AI is doing with music.

As you can imagine, this is a controversial area on the copyright and IP front, and one of the tools I’m showing you today did start out allowing you to create music in the style of a given artist and it raised such a furore (quite rightly!) that guardrails went in.

I tried the following prompt: A Robbie Williams song about lonely travellers on a train across France.

• Suno just straight up wouldn’t generate anything.
• Udio gave me an error message saying “we do not generate artists’ likenesses without permission” then listed the style tags it had replaced his name with.
• GoogleLabs’ Music FX told me it couldn’t generate anything from my prompt.

So at least this is one area where some progress is being made around artists’ rights.

With that heavy stuff out of the way, here’s a prompt that DID work: A happy pop song about a cute little robot who likes to learn about AI tools and fun ways to use them.

Suno works by creating lyrics in ChatGPT, then setting them to music. You get two full song versions, both with the same lyrics. You can also input your own lyrics which can be fun if you want to include something specific.

Udio’s music tends to be more complex, and the lyrics vary across the two versions it produces. But you only get 30(ish) seconds, then need to generate again to extend it, which doesn’t always go brilliantly. Again, you have the option to include your own lyrics, but you only get about six lines.

And finally GoogleLabs MusicFX. I haven’t played with this one much, I find the other two a bit more consistent in their output styles and quality, and I like their interfaces a bit better. Although this tool can be fun as its focus is on originality – I’m pretty sure the lyrics in both these songs are in Japanese, which was a bit unexpected, but then, I didn’t actually ask for English…

A tool I haven’t sampled here, but which is quite popular for backing soundtracks for YouTubers and other video creators is Mubert, which only does instrumental pieces (they describe it as royalty-free background music). You can choose the length as well as your text prompt, and it does have an audio watermark – randomly saying ‘Mubert’ at some point in your track – which you can get rid of by buying it in the same way you would a Getty image.

All of these tools have a free tier which works on a credit system, similar to the Image Generators.

• You get 50 credits a day on Suno, which is enough for 10 songs. They don’t roll over, but reset each day.
• Udio gives you 10 generations a day (up to 100 a month, so you can’t go in every single day and do 10 of them).
• MusicFX – no limits documented, I’m not about to try…
• Mubert – 25 tracks per month.

As with the image generators, anything created on these free tiers are only intended for personal and non-commercial use.

Have fun!