Is AI Music Easy to Create?

There is a joke to be made there about artists who are a bit zero-skilled already on streaming platforms like Spotify. But we’re looking at how easy it is to generate a song and get it published on DSPs, with zero talent or musical skill.

And what better way to explain it than to show you what we did and share the output (music)?

For this, we’re using Suno’s Pro Plan (which includes commercial use rights) and Ditto Music. There are a couple of things to be aware of: not all platforms will accept the AI track - it depends on their policies. Another thing to know is that if you are using other people's lyrics or parts of their song, you don’t have 100% exclusive rights to them. And, which we learned the hard way, signing up for a pro plan on Suno AFTER generating music doesn’t give you a retroactive license, and so nothing there can be monetized. Drat.

Also, YouTube might be off-limits to you due to ContentID, and yes, that includes YouTube Shorts. BUT you have 159 other platforms that might accept your music, so the odds are in your favor of going live somewhere (clicking the first two links below will take you to the song and the Udio platform).

Initially, we tested out Udio. Here are two tracks generated by different members of the team (without discussion of how we were using the tool):

Sample1
Prompt used: Suspenseful piece for opening a bag of chips too loudly in public, cinematic, horror

Sample2
Prompt used: A melancholic, atmospheric track featuring bittersweet, gravelly male vocals with heavy reverb and subtle "telephone" filtering. The production should feel lo-fi and "crusty"—think dusty vinyl crackle, a deep, side-chained sub-bass, and a slow, purposeful trip-hop drum beat. Incorporate a sad, repetitive piano melody and lush, sweeping string pads that swell during the chorus. The bridge should feature distorted vocal layering and a sense of "beautiful chaos", experimental, synth-pop, Lo-fi R&B, Trip-hop, Neo-soul, Experimental Pop, Cinematic Sadcore.

It’s not hard to see a huge difference in how they sound. Obviously, the prompts had a significant impact here. But you can be as lazy as you want, type the style of music you’d like to create into Gemini or Claude, and ask it to generate a prompt of around 1000 characters.

Then we ran into an issue (after using all of the free credits to make a bunch of tracks). Udio doesn’t have an in-built download option anymore. You can play around and publish them on the platform, but the downloading of audio, video, and stems has been disabled. Annoying, but that’s what you get when you don’t read the FAQs first.

Next option, Suno.

The platforms are almost a copy-and-paste of each other, so it doesn’t take long to get to grips with them. This time we can download. Nothing can stop us now…

We know that AI music from The Velvet Sundown and Breaking Rust have both done exceptionally well. Breaking records and stacking the revenue high in 2025. And we also know that most people who use streaming services are ‘laid-back listeners’. Vocals can be a giveaway on AI tracks, sounding robotic or thin, so we aimed to use fewer sections with demanding vocals. The styles most likely to trickle into playlists are lo-fi, chillbeats, chillhop, and lo-fi trip-hop, and anything that fits neatly into those genres should be a green light. They all sound the same after 20 minutes anyway.

Giving ourselves the best chance of making our way, with zero skills, to the masses. With a commercial license and some good prompts, we’re on our way.

Should it be this easy to create music? And, just because it sounds like music, are we going to classify it as music?

We know that 97% of people can’t tell the difference between what is AI music and what isn’t. Even we can’t tell anymore.

So, with a revised prompt, a new platform, and a bit of time (please note clicking play will take you to the web version), this is what we ‘created’:

So now we are starting to produce sounds similar to the millions of lo-fi tracks on so many platforms. It seems like low-hanging fruit, but the process is so easy that we’re starting to wonder why more people don’t do it, rather than hoping everyone stops.

When you’re listening to something that has no vocals, does being human matter? Murky waters.

We mentioned we’re using Ditto Music to publish, and we have. Though we didn’t want to stump up the extra £40/$40 for a 3-day release, we’re now in a free 10-day queue. But that gives you plenty of time to use their pre-release bundles and get people excited about whatever you’re releasing.

Was it difficult? No, it was fast and easy. Ideally, we’d have known there was a 10-day until release timer, but we’ll update you with actual live links when we have them, and follow what the published side looks like too. For this, we will use a combination of Ditto Music (streaming data) and Music24 for playlist add analytics.

For a very reasonable $59 for a year, you can get unlimited releases for two artists on Ditto Music. There are other perks too, but if you want to keep it even simpler, it’s $19 for a starter plan. But we also know that under 1000 plays don’t get paid (88% of tracks never break the barrier), and there are something like 27-50k songs uploaded every day. The chances of seeing a return on that investment are very small, but not impossible. And then, if the experiment works out, we will be able to report any income. An interesting experiment to see roughly how long, if ever, you can recoup your cash (these tracks included here don’t count and won’t be released on any streaming platforms). And, if you can get paid anything at all for just a few hours’ work.

We’re not advertising any of these platforms; we're just sharing the process.

So what about something different? Not lofi, but something with lyrics and a different genre?

If we hadn’t introduced the newsletter as AI, would you have known that the next song was AI-generated? And, if you enjoy it, would it even matter?

We ran it through SubmitHub’s AI checker. Here is the folded napkins score:

Interesting results. Because we all know these are 100% AI-generated, but to someone else listening, would they be able to tell if an AI checker can’t tell? Feel free to test the theory. To the casual ear, or even a trained one, we are at the point that AI-gen music is fooling the best of us.

In the end, our experiment is ongoing, so we don’t have a hard conclusion yet. But with a few hours and a minimal financial investment, there is barely any distance between having an idea and achieving it. It doesn’t matter if we call it art, music, or just utility audio anymore. Streaming platforms have very few guidelines for what will and won’t be accepted. The true gatekeepers are the distribution platforms, who can review and reject at will.

At no point did we need to feel anything to make what you’ve heard here. The tester track that is going out into the world just needs to be sorted, slotted into an algorithm, and fed to listeners. Platforms are moving away from ‘delete and ban’ toward disclosure; the floodgates are now wide open.

We’ll let you know when we’re live and if our zero-effort experiment generates a single cent.

Until next time,