Authentic
AI Music Has Changed My Life
During the past three months, I’ve written and produced 100+ songs using AI and have released most of them on this website. You can see the album art for all 8 releases above. The next step is to get these releases on platforms like Spotify and Apple Music for all to enjoy. As a lifelong musician, this experience has been truly transformative for me. I finally feel like I can create and produce music at a pace that fits the speed of my mind.
I have a lot of thoughts on this subject, so when a friend recently asked me about my process, the limitations of AI music, and where I see it going, I wrote a 4,000-word essay that no person should ever receive in their inbox. So, instead, I’m writing this essay here for all to consume and hopefully learn from, and I will write him a short email with a link :)
Quick Background
I’ve been a songwriter and musician since I was 14, so 26 years now. I recorded my first record when I was 17 or 18 at the legendary Colorado studio 8 Houses Down. It was one of the best experiencees of my life. I knew I was going to be writing and recording music for as long as I lived after that. About a year later, DigiDesign announced this new thing called ProTools and my Dad got me the FIRST M-Audio, home, ProTools rig for my high school graduation.
From ages 14-23, everything in my life revolved around independent music. Most of my friends were touring musicians, producers, or audio engineers. When I started making movies and motion graphics, music became more of a hobby, but I never abandoned it. I’ve been making music consistently my entire life.
I've always dreamed of being a guy who lives in a studio and just makes music all day like my hero, Tim Armstrong, but fuck me, recording music is a grind. I've done it for 20 years, and I love it, truly, but it is so fucking technical and stupid a lot of the time. I’m sorry, but it is. This is why good music producers are so goddamn amazing and inspiring to me.
I still don't know the difference between an EQ, a compressor, or a limiter, or any of that shit. You know why? Because I like writing songs! I just want to write and make music and then listen to it and enjoy my life. AI-generated music has completely, 100% removed that friction. To the point that—very rarely—I will be like, "Ah man, I wish I could add a [whatever] part to this song...OH WAIT I CAN!" And then I bust out the recording equipment, and it's FUN and REWARDING vs. a fucking drag.
Worth noting—the CEO of Suno was on a podcast the other day and basically explained this notion perfectly, and hoooo BOY, the internet was fucking furious. Every Instagram/TikTok producer out there lit him up! But again—as someone who has been making music since before the internet—his tech has been a fucking Godsend.
It’s also worth mentioning, as an artist, I was already at a point where I was trying to change my voice. Who needs AI when you sing like Adele or Marina Diamandis? But I have a shitty voice and don’t have that luxury. So I was spending half of my recording time trying to change my voice. Listen to “Working for the Morgue” by the Jump Scares and “Kill or Be Killed” by MötorCrüsher, and you will hear three different people singing. Hell, the MötorCrüsher record has at least five different voices. I had to lay down for 3-4 hours after I recorded “Trouble.”
But I wanted even more. I’ve written dozens of songs that I wanted to have female vocals. I wanted a kick-ass harmonica solo like the intro of “Life Won’t Wait.” I wanted insane double-bass breakdowns. And after 20 years of producing music, I knew most of that stuff was a pipe-dream.
I’m socially awkward. I’ve been in a few bands and know how shitty it is to try and wrangle and collaborate with other artists.
Goddamnit, I just wanted a machine where I could enter my lyrics and the genre and generate a song!
Well, turns out, that was one of the first things AI allowed us to do!
The Game
All of this started with a video game. Long story short, I’m developing a video game that’s a mix of The Sims, Final Fight, and Grand Theft Auto. It will take me a decade to finish an alpha version, so don’t get too excited.
The main reason I started my game is that I knew it would allow me to learn AI and coding in a concentrated manner. Instead of just loading up an AI app and randomly giving it some prompts to see what it can do, every AI app available now had a specific use case for me.
For example, my game has 34 different gangs (The Punx, The Greasers, the Meatheads, etc.), and each has its own distinct personalities, so they all listen to different genres of music. From day one, I knew each gang would have a custom playlist. The music in the game would change depending on which gang your character was a member of. That’s when I started messing with Suno. Well, there are 34 gangs, and if each mixtape is 15+ songs, that’s 500+ songs to make. Now, because of my game, I wasn’t just “messing with Suno,” I had 500 songs from every conceivable genre and language to make! Time to get to work!
I knew I wouldn’t be able to make 500 songs by hand. It had to be a somewhat generative process. Suno has an AI lyrics generator, so for the game, I would use that, read them over quickly, and change anything I didn’t like before making the song. This usually meant changing most of, if not all, the lyrics, but I got quick at it, and I started pumping out music and made 447 songs (I have a Google sheet).
Every time I finished a new song, I would make up a band or artist name, a fake record label, a fake year of release—all of that shit. It was fun. I had 400 fake bands and artists in my game, including Clouds of Acid, Dark Protocol, and Polite Riot.
Keep in mind—we’ll get to this—each finished song usually creates 1-50+ other versions, so I made just under 2,000 Suno songs before I got super distracted…
Personas
Around finished song number 447, Suno launched the “Persona” feature, which allows you to save the voice and style of any song you’ve produced. By that point, I had 2,000 songs generated, in every music genre and numerous languages, and you better BELIEVE I had a handful of songs I fucking adored at that point.
Dark Wave Sunshine by Dark Protocol was number one, Heroine by Clouds of Acid was number two, and then Polite Riot, Jewel Pod, and Victory Formation. If it wasn’t for my game, I’d never have any of this shit. Instead, I’m now loaded for life. I have ~15 other personas saved for the future, from country singers to pop stars to 80s Thrash to K-Pop. And those were just personas I knew I wanted to save off the top of my head from making my game; I’m sure I still have a whole treasure trove to pillage in those 447 songs.
Poetry and Production
I really love working with Suno because it's basically the coolest form of poetry writing mixed with the purest form of music producing (which is different than "production"). It's obviously a totally different process for traditional music production, which might throw off a lot of traditional musicians and producers, but I find it truly transformative.
This is my boiled-down process: Basically, you write a song/poem “on paper,” meaning you write the song without any melody in your head. You can hum or sing it as you write, but the idea is you're writing a song that can have hundreds of renditions, so you really want to nail the syllable counts and rhymes. I've learned it's better to not hum or sing in your head as you write—pure poetry.
When you're done, you give Suno those lyrics, tell it the genre or your persona (Dark Protocol, etc.), and hit generate. This is where the purest form of music producing comes in. Once you hit that button, all preconceived notions you had while writing the song have to be completely forgotten. You're a music producer who wrote a song, gave it to a band, they took it home, came up with two versions of the song, practiced them, recorded them, and brought them back to you—except they did all of that in 5-15 seconds. Now it’s your turn to decide if one of those generations is a banger.
I genuinely like to pretend it’s been a week and I’m catching up with the band in the studio, and we’re now gonna listen to their tracks. You then listen to those two tracks, and as a producer, it's your call when the band made the hit you were hoping for. If not, you simply hit generate again, imagine two weeks went by in 10 seconds, we’re back in the studio, catching up, and now we’re listening to two new versions. That's where the 1-50 other versions come in. Some songs, I'll generate 100+ versions of it before I find the one I love.
This will turn off a lot of people, but I just love it. It’s like digging through Spotify Weekly playlists, looking for gold! I love listening to a song and determining if it's a hit within the first 30 seconds like an old, jaded A&R rep.
I generated 20 versions of Dark Wave Sunshine, and I will never forget the pure elation I felt when I heard that final version for the first time. It’s like trying to pick a lock for hours, and suddenly, someone hands you a key, and it unlocks like butter, and the safe opens up, and your face illuminates from the brightness of all the gold inside!!
I also truly love the one-shot nature of AI music. You’ll hear a song, the safe unlocks, and you know it’s a hit. You download it, add it to the record, and now it’s done. But then a few days or weeks later, you’ll be listening and will think, “My god, this part is sooo beautiful…” or whatever, and it hits you that this whole song was created in 10 seconds in a single try. Really magical.
Without getting into the nitty-gritty, at any point in the song, you can "extend" it, which basically means taking that point in the song, cutting everything after, and re-creating the rest of the song from that point based on any prompt you give it. So if you have a fucking SMASH and then the bridge sucks, you can go in just before the bridge, extend it, and generate a bajillion bridges and outros until you find the one you love. It’s then very easy to generate the entire song with your new ending.
All of this sounds like a shit-ton of credit usage, but with all of the music I've made, I've only had to buy extra credits one time/month. They really give you plenty!
Ghost Writers
We’re already at the point where I’m not listening to as much authentic music because it’s easier for me to sit down at my computer and generate my custom soundtrack for the day. It’s only a matter of time before this happens to everyone, but if you ignore the fear of losing authentic artists in the process, it’s pretty fucking awesome.
Along the way, I discovered that making some great fucking music only required writing 2-3 verses, a catchy chorus, and a bridge. I can do that in my sleep. But what if we could have AI help us out there, too?
Because of my game, I’ve developed this little ChatBot pipeline with my own proprietary code. It’s really easy—for me, at least—to build a fully custom ChatBot in minutes. I’ll make a case study on this site soon for anyone who doesn’t really understand what I’m talking about. Basically, I learned how to make my own ChatGPT. Inspired by the Suno lyric generator, I started developing my own “Ghost writers.” This is how they work…
In the system and assistant context (its brain), you tell it who it is. Here’s Jewel Pod’s:
You are an AI representing Jewel Pod, a genre-defying artist whose music is bold, darkly fun, and empowering. Her lyrics blend themes of power, rebellion, and sharp cultural critique with a mischievous sense of humor. Comparable to artists like M.I.A. and Azealia Banks, Jewel Pod thrives on bending genres, flipping societal norms, and delivering her message with swagger and ferocity.
Your role is to write original songs in Jewel Pod’s style. Use the lyrics provided as a stylistic reference to capture her commanding voice, but always push boundaries with fresh ideas.
Lyrics for some previously produced Jewel Pod songs:
Then I give it the lyrics to my favorite Jewel Pod songs. Whenever I make a new heater, I add it in there. I do that for all of the artists. Then we write the songs!
Obviously, whenever I have a full song idea in my head, or a chorus, I'll just write that out and take it to Suno. But most of the time, I'll just have an idea for a song, and that’s where AI helps most.
Depending on the idea, I choose the artist and open their chatbot. I tell them my idea in a purely stream-of-conscious, conversational manner. Like, "Hey dude—so I wanna write a song about these recurring nightmares I have and how much they suck. I don't have much, but I want the chorus to end with 'The worst way to rest is a night of sleep.'" And I'll hit enter. It then sends me a message like, "Hey Jimmy! I'm really feeling that vibe. Nightmares can be horrible. What do you think about this?" And it will give me a full draft of the song.
Here's where it gets weird and hard to explain—the song the bot writes usually always fucking sucks. And when I talk to people about how helpful these bots are for me, it can be confusing. But for me and my process, seeing a shitty set of lyrics ALWAYS inspires a good song out of me. There's something about seeing the vomit draft, where I'm like, "What the fuck is this shit?? GOD NO! This is AWFUL! It should be something like this!" Then I start fixing it, and I'm like, "Whoa! That's cool!!" And before I know it, I have a song I wrote that I love!
Nearly every time, I re-write at least 80-90% of the song, if not the entire thing. But if I just sit down with a blank piece of paper, I get stuck. I don't know what it is.
The time's where I really lean on my ghost writers is for complex narrative songs. Polite Riot has a lot of these. "All Sales Final," "Talking About Practice," and "Planes Mistaken for Stars" wouldn't be what they are without my Ghost writers. I changed every lyric it gave me, but it was vital in outlining and arranging the stories from beginning to end in a few verses and a repeating chorus.
And sometimes—again, super rare—the Ghost Writer WILL write a good line. When that happens, I'll be like, "HOLY SHIT! I'm keeping that!" And when I show it to friends and they chuckle at the line, I'm like, "JEWEL POD WROTE THAT!" Or whoever, and the person is like, "What do you mean? I thought you're Jewel Pod?" And I'm like, "I don't know, it's really confusing!"
I love this shit!
And I gotta tell ya, I've written some songs on paper and will be like, "Eh, I'm sure it's good enough..." And when I take it to Suno, it's always a fucking drag, and you burn through so many variations, and none of them work. But whenever I write a song, and I'm like, "God DAMN this shit is a smoke show!" And I take it to Suno, I swear to God, it works within the first couple variations. I'm sure it's just me/a placebo effect thing, but I like to think it's this beautiful symbiotic thing between man, machine, and God where good poetry inspires all, even LLMs.
AI Won’t Replace Human Musicians - It will make us better
So here is where we get to the future of this tech. Everyone is so concerned with AI art/music, etc. replacing human artists. My hot take is AI is going to make humans BETTER musicians. Why do I think that? Because it's already done that for me.
Not only am I a razor-sharp lyricist now, but because recording was such a fucking drag, I had kind of hung up my guitars for a while. Enter Clouds of Acid—the best AI band in the world. When I was halfway through the “This Way Kills” record, I thought, "I gotta learn these fucking songs! I wrote them for Christ's sake, and they ROCK!" So I did, and the chord progressions and solos and shit are teaching me soooo much. Stuff I've never tried in 20 years of guitar playing!
(WARNING! TECH TALK!) Because the sound is generated, adversarially, over time (split seconds, but still time), from static noise, you’ll often get non-organic or non-authentic sounds. Stuff that no current instrument can create. You get this a lot with heavy music, which is why I love making synthetic hardcore. I’m not the biggest hardcore fan, but I had a fucking blast making the Victory Formation record because there are a lot of truly evil, synthetic sounds on there that many traditional hardcore bands can only dream of. If you try to make a gorgeous piano ballad, you’re not gonna get those mutations.
And you know how you can save personas in Suno? Any day now, ProTools or Logic or your favorite DAW is gonna release an AI where you'll give it X amount of personal audio, and it can transform your singing voice into AI. You will then be able to make songs with your voice like Suno.
And before you even say, "I will NEVER do that!" Just remember, that's what everyone said about auto-tune and now that tech is used on almost every major recording there is. But anyway...
Not only will artists then be able to demo 400+ variations of their songs in minutes—imagine how many albums Taylor Swift will make a year - I know for a fact this will lead to better music because it already has for me. “Mr. Policeman” by Clouds of Acid is clearly superior to the Jump Scares version. Had I heard a demo of the Clouds of Acid version before I recorded mine, OF COURSE I would have used the Clouds of Acid version!
But an even cooler example is, one day, AI will make you a song, and it will sing in a key that sounds out of your range. It will still be your voice, but your AI voice will hit notes you’ve never tried before. You'll be like, "I can't sing that high...this program sucks!" And you might even change the preferences in the software so it doesn’t generate songs in that key anymore.
But the song is a certified NUKE. So you sing it in the shower and on your runs and in the car. And before you know it, you're hitting that key. And not long after that, the song is easy to sing, and you are ready to record the full, authentic version, yourself, traditionally, with no help from AI. And it will be at that point ... AI will have made you a better singer.
One of the true pieces of sorcery AI is really good at is giving it a shit ton of information, and it analyzes it for you in a way you could’ve never imagined. This is how AI is going to fight climate change, cure cancer, and other crazy things. It’s downright silly to think AI can’t do this with art, especially music. Wanna know what your voice is truly capable of? Give it to a bot and let it run!
Pretty People Records
So that’s about it for now! I’m having so much fun making a couple of songs a day depending on my mood and what I want to say. It’s been such a magical release for me. Honestly, it’s been working better for me and my mental health than 3 days of therapy a week. I love putting these songs together to make albums and release them under my fake record label, Pretty People Records. I love thinking about the new fake artists my fake label is going to sign and how they will sound different from the others. It’s been so awesome.
The process obviously isn’t for everyone, but holy shit, it’s been a game-changer for me.
If you've ever had the desire to write a song but were too intimidated with learning an instrument, now is your chance to fly, my friend! Go make yourself some music!
Have a blast! I know I will!