Unlock Creative Flow: Create and Distribute Music with AI

Full video transcript

Hey, how’s it going? My name is Darby Rollins. I am the founder and CEO of Gen AI University, and in this session I’m going to be sharing with you a practice that I call productive play as it relates to using AI for enhancing creativity and having a little bit of fun learning new skills in the process.

What I’m going to be covering inside of this video is going to apply if you’re interested in music, creating sounds, songs, and just experimenting with AI tools that can help bring an idea or a vision that you have for a piece of music to life. Keep in mind that prior to this session you may have no experience with music, you may already be an experienced musician, or you may be considering picking up music but you just don’t know where to start. The great thing about where AI fits into this picture is it doesn’t matter what stage of being a music artist, or just experimenting with AI in music, that you’re at right now — what I’m going to share with you applies to improving your skills on your own and flexing those creative muscles along the way.

Throughout this video I’m going to show you a few very specific tools that I’ve used personally when practicing my own version of productive play to create music. Being a drummer and being involved in music, I’ve had a lot more exposure to music than someone who is just getting into creating music right now. But when I started using AI to help with the process of creating music, it solved a very specific problem I was experiencing, and I wanted to apply myself through this process of productive play to learn not just how to create music, but how to flesh out ideas for songs that were in my head and actually get them out and distributed to the world.

I was stuck in terms of my own experience: I have a drum set, an electric drum kit, and I lived in an apartment for the majority of the past few years (I’m living in a house right now). I love playing drums and playing along with music tracks, and I know that drumming stimulates my thinking, gets me into flow, and is a way for me to sort of meditate with myself. There are a lot of constructive aspects to learning and playing music that you can read about with a simple Google search or by asking ChatGPT to walk you through how music can be productive for developing your brain and additional soft skills.

Throughout this video I’m going to show you the process I’ve gone through — not only ideating a song but actually working with an AI platform to break through writer’s block for musicians. I’ve seen in comments and on Reddit that other musicians using AI tools like this in their workflow say it’s not replacing the act of being a musician for professionals, but acting as a tool to stimulate the thought process when you run into a music block that slows down creating and producing.

The tools I’ll walk you through show how to start from scratch with a few simple prompts describing in your own words what you want a song to sound like, then how to iterate on those inputs to refine the outputs, and ultimately create what I’d call about 80% of a finished song at this stage — enough to then fine-tune the rest in a simple tool. I used GarageBand for this because I have a Mac and it’s a simple, free way to get started, but the same workflow applies with other tools.

Where I wanted to learn how to use AI in this process was not only to create a song but to create a song I could distribute on platforms like Spotify, Apple, YouTube, and other streaming platforms. Knowing how to create music and art is amazing, but if you can’t get your creations out there and distributed across multiple platforms so people can easily consume your content, there’s a lot of opportunity going untapped — it takes systems and the right levers to pull, and obviously you want to be producing content that the people you’re making it for actually want to listen to.

This ties back into productive play and where I’m going to start the demonstration: producing for an audience of one. When we’re producing something with a desired outcome for other people — a homework assignment due today, or a project for a client with a specific deliverable — using AI for that can feel like work, even if you’re doing it faster and more efficiently, because your brain stays in work mode. I like to explore AI tools that aren’t tied to a client project or a deadline, so I can have fun with them and still be productive. That’s what I’d encourage if you’re getting started with music: create for an audience of one. In my case that’s me, or it could be someone you want to produce for — I make songs using these tools just for my fiancée, with no intention of sharing them outside that, because I know the type of music she likes and can use her input.

What you’ll see as an end result is a song I produced — my second song, technically, but the one I’d call a Rock Anthem I created for myself, called “Keep Calm and AI On” by Donnie Hammer, an AI-powered music artist alias I set up with social media channels and distribution. I wanted to figure out how to go from an idea for a song, to creating it with AI, to editing it a bit in a tool like GarageBand, to publishing it and getting it distributed so I could open Spotify and listen to my own songs.

Let’s dive in. We’re looking at the library in a platform called Suno, a generative AI music creation platform that lets you write prompts and extend specific pieces of sound and music to create full songs. The process is simple: when you click create in Suno, you can write a song description, choose whether you want it instrumental, and get ideas from Suno itself. On the right you can see a lot of different songs that have been produced or started, using prompts that describe what you want to hear.

For example, when creating from scratch, you put in a prompt like “rock music” with instructions describing your favorite sounds, styles, whether you want a male or female vocalist, and any lyrics. When you create an initial output, you get a one-to-four-minute clip that you can then extend from specific sections. If you liked the first 30 seconds but it went in a direction you didn’t want after that, you can extend from that timestamp and create different variations by refining your prompts. This is exactly how “Keep Calm and AI On” was created — I had a title and wanted to find a good intro to build on, and the entire process was finding those clips and extending from the pieces I liked to reach a fully fleshed-out version.

Going back into the library, you can see part one and its different revisions. The prompt for part one was “a new metal rock song in 140 BPM with cosmic new metal sound for riffs telling a story through time and space, mantra house” — mixing a few genres and sounds I like. You don’t know exactly how it will sound until the prompt articulates those sounds properly. You can do this for nearly any type of music, from a solo flute or acoustic guitar to combining multiple instruments — it’s a matter of describing what you want to hear and how fast the tempo is.

I ran through a number of variations based on the outputs, and once I liked an output I would extend it — you can see part two, part three. This let me take a one- or two-minute clip and extend it from any section I wanted, based on my preference for how I wanted the song to sound. Every batch of credits (in this case roughly two outputs per five credits) gives you many different songs to experiment with, which helps break through writer’s block even if you don’t know exactly what you want going in. I probably spent an hour or two playing around and stitching this particular song together across three parts, ending at 3 minutes and 11 seconds, with a final 14-second closing track I generated four or five times to get the ending I wanted.

Once I had those pieces, I moved on to cleaning things up a bit in GarageBand — I’m going to skip the details of that since there’s a lot of information about GarageBand already on YouTube. The important thing is having the full clip of the song you want to produce; you can share the clip directly or download it as an audio file, which I’d recommend for the next stages, including mastering it for sound quality.

The song I produced was pretty much 100% AI besides my direction iterating on the process, but a lot of musicians using these tools are using them as an ideation and starting block, then layering in their own playing and other musicians on top.

From a finished clip inside Suno, the next step for me was DistroKid. Like a number of platforms, the goal is getting your music onto Spotify and into people’s ears. You’ll also need things like album art, which I’ll show how to create quickly in ChatGPT. I’m not a professional music artist doing this full time — I was doing this to have fun as a hobby, and you’ll see a lot of different offers and services on a platform like DistroKid; this isn’t a blanket endorsement of everything there, just a portal for getting your music out into the world if that’s what you want.

To upload into DistroKid you set up a personal profile, payment methods, and answer some basic questions — you can add your artist or band name and set up distribution to Spotify, Apple Music, YouTube, and more (some of these can help you set up your YouTube Music presence too). You’ll need an album cover, which you can create yourself, source from free image sites, or generate with a tool like ChatGPT, which is what I did. You upload the audio track, and one thing you’ll run into is mastering the file — a pretty standard step that helps the song sound better on speakers, with both free and paid mastering options available.

After uploading, your content gets submitted and distributed over the course of about a week or two, and you can track stats on how many people have listened. Over the last 365 days I’ve had 42 listens, and most of the last seven days I can pretty much guarantee were me playing along with drums. DistroKid also lets you add splits for collaborators if you’re working with other artists, which is useful if you want to co-create and get in front of new audiences — collaboration and distribution go hand in hand for growing a music career.

The last piece is album art. Both album covers shown were created in ChatGPT, then I took those images into Canva to add the title text. In ChatGPT, you describe what you’re looking for in natural language — for example, “creating an album for a song called Keep Calm and AI On, it should be cool and futuristic and be related to drumming.” You can iterate from there, like asking for more drums since I’m a drummer, and continue refining the image. You can also use ChatGPT to create additional posts to help promote your song once it’s finished.

For the final artwork, DistroKid recommends a 3000 by 3000 pixel image. I downloaded the ChatGPT image, uploaded it into Canva, and added a title, adjusting position and transparency as needed — you don’t have to add anything if you don’t want to. Distribution platforms have guidelines around what can appear on album art, such as not including a domain name, so keep that in mind. Canva is free for most of what you need to get started, with paid tiers for professional features.

Once the artwork is set, you go back into DistroKid, submit your content, and your song — from a Suno prompt and your own creative direction — ends up published on Spotify. To wrap up: add your album artwork, select the remaining distribution options, and continue; within a matter of a week to a few weeks your song appears on the popular streaming platforms. This all started with an idea for a prompt and a song we wanted to create, done in the context of productive play for an audience of one.

Now that I know the entire process, going back into Suno I can see there are so many different types of music — meditative, EDM, trance, tribal, new wave — to experiment with for inspiration. There’s a lot more that goes into being successful in music than just putting songs on Spotify, but you have to be able to get there first, and now I know how to create, ideate, and mix those songs together for my own playlist, which could grow into more one day.

That’s the practice of productive play as I use it in my own music, and I hope this process has been informative and helpful in showing you from start to finish how you can leverage AI to create and produce music — starting with a tool like Suno, mixing in a little ChatGPT, and using platforms like DistroKid to distribute your music to popular streaming platforms like Spotify, Apple, and YouTube, and how you can potentially even monetize your songs as you grow as a musician. More importantly, I hope you found this process of productive play useful for learning to produce music and having a little fun along the way.

If you’d like to reach out and connect with me, feel free to find me online, on LinkedIn at Darby Rollins, and let me know if you put this into practice how productive play works for you. I look forward to hearing how you use this concept to create some cool music, and as always, keep calm and AI on. I’ll see you on the other side. Cheers.