Over the weekend, I was playing around with Suno, the AI music generator, having read the article about it in The Verge.

I was inspired by the article to write a country song and see what it could do with it, and the results (especially after refining by making a second version) were pretty remarkable. Embed, with Suno AI-generated image included:

That got me thinking — if Suno could take my lyrics that were thought up in about ten minutes and turn it into a song that could be on country music radio with a little bit of touching up, what could it do for real country musicians?

The article has a clip of a song by some songwriters that’s just vocals and guitar on a voice memo — no mastering or anything, side-by-side with the same clip, remixed by Suno, which keeps the melody and rhythm, but makes the song radio-ready.

With this new technology where pretty much anyone can make any songs they want with lyrics and a prompt, professional musicians are going to be forced to be better than Suno. Beating users on the free plan is easy — v4.5-all, their free model, isn’t going to match the creativity of a real artist. However, their paid option, v5, is more accurate and less predictable than v4.5-all.

Suno’s not going to get rid of musicians yet, or at all. For celebrities, part of why people are listening is because they want to listen to the celebrity — or at least at this moment — because music, since its very start, hasn’t just been about the sound. It’s about who’s singing, why they’re singing it, and how it makes people feel. Suno is highly impressive, and I’ve been blown away with the quality of the songs I’ve created. But musicians, for now, have creativity, something that AI doesn’t have. Artists can decide if their song needs a trumpet or xylophone solo and make their choice for a reason. AI can make the choice, but even if it pretends it has a reason, it might’ve still made the wrong decision.

Where Suno really shines is empowering budding songwriters with the tools to create great tracks from their own music. As the article explained, musicians previously needed to pay somewhere between $500-1000 to get a “studio guy” to record a demo track to pitch to a publisher. Now, they can just use Suno, and get a pretty much equal result in thirty seconds. These demo tracks are never meant to be heard. They are just to send musicians to record. These demo recording artists aren’t making those same choices that the artists can make once they received the song — they are just there to serve a role in the process. That these people are being replaced by AI has no impact on the final creativity of the song. It makes it easier for songwriters to do more at once, and, if anything, might result in a positive change in the music industry — more songs from skilled songwriters.

It also allows non-musicians to dip their toes into creating music — something that used to have a very high bar to entry if you wanted a professional-sounding song. Now someone who’s never recorded a song before can just type lyrics into Suno and a prompt for what they want the song to sound like, and they’ll have a song. Suno lets people not have to hire instrumentalists to make a song for fun.

I think that Suno threatens the “track guys” only. For songwriters, it allows them to do more. For the session players, a real musician would hire them if they want their song to sound even better. And it lets normal people create songs for free, which opens up the music industry to more people.

I feel like they could charge a lot more. If the article in The Verge is true, that every Nashville songwriter is using it, they could charge more than $72 a year for 500 songs with their v5 model, under their Pro plan.

Overall, I’m super intrigued by Suno, and am thinking about upgrading to the Pro plan mainly for access to v5 and uploads longer than one minute. It makes the whole process of creating a song a whole lot less intimidating for beginners like me, but it’s also really interesting to see just how much it can do. To me, Suno feels like one of the best uses of AI I’ve ever seen. Using AI to write or to code doesn’t leave you walking away with something that you feel is yours. When you use Suno and put in your own recording for it to make it better, you get something that you made — just amplified to a level higher than you’d ever be able to achieve.

I hope that Spotify doesn’t get overrun with Suno artists, but at the same time, I know it will. With the power that it gives the public to create music, we’ll have to adjust, just like we’ve done to detect when a piece of text or an image is AI-generated: we’ll have to filter through the noise and find artists who are making music because they’re passionate about it. Right now, no one is really suspicious of music being AI-generated in the same way we are of videos, images, and text. But I think we’ll soon have to be. I’ll close with an amusing comment on one of my previous posts from Hacker News that feels relevant here:

I didn’t use AI — all of my words are mine — and so are all of my em dashes — just not my song — I mean, sort of — I wrote the lyrics.