Luke F. Walton Love Music More Episodes BBL Drizzy and Burrito Bot (AI Song Generation Tips)

BBL Drizzy and Burrito Bot (AI Song Generation Tips)

Love Music More · hosted by Luke F. Walton · Solo episode

The podcast feed is titled Love Music More with Scoobert Doobert. Scoobert Doobert is Luke F. Walton's primary music project.

Jump to section
  1. Listen
  2. Topics discussed
  3. Host note
  4. Selected moments
  5. Selected excerpts
  6. FAQ
  7. Transcript

Listen

Sponsor the show.

Topics discussed

Host note

"I don't want a sugar daddy, I want an enchilada mama", that line sparked joy the moment it came out, and it's also a pretty good summary of what makes AI music generation interesting when you stop trying to make it sound serious. Burrito Bot, my side project dedicated entirely to songs about Mexican food, turned into a genuine laboratory for figuring out where these tools break down and where they don't.

The thing that really unlocked it was writing my own lyrics rather than letting the AI do it. AI lyric generation is still the weakest link, but plugging that gap with human taste is what makes the whole thing work. BBL Drizzy is the more high-profile example: a voice transform done with AI, anonymously, that went viral. The tools are already in use; the question is how.

A lot of the artistic magic is in the crapiness of it. I'm not looking for perfect from these tools. I'm looking for the thing that only happens because the machine doesn't quite know what it's doing.

The essay is I Made 1000 AI Music Songs So You Don't Have to, the same day as this episode. The records are Refried Being and Wrap Music.

Selected moments

  • Introduction to AI Music Project 1:31 I explain my AI music project and the unique genre I i creating.
  • Burrito Bot Overview 2:19 I introduce side project, Burrito Bot, dedicated to writing songs about Mexican food.
  • AI Lyric Generation Challenges 3:02 I discuss AI's limitations in crafting lyrics and the importance of human touch.
  • Example of AI Song "Heart On My Sleeve" 4:34 I reference an AI-generated hit as an example of voice transformation in music.
  • Human Involvement in AI Music 7:32 I explain the importance of human creativity in the AI music-making process.
  • Conclusion and Future of AI in Music 11:17 I share my thoughts on the ongoing evolution of AI tools and their impact on music.

Selected excerpts

I'm trying to make it. And a new little bit was discovered and uncovered with suNo.ai... it turned it into something that I feel is now, finally, listenable.

~1:38 in the full interview

But the thing that really unlocked it was when I was able to start writing my own lyrics for it.

~2:14 in the full interview

I don't want a sugar daddy, I want an enchilada mama, and that just sparked some joy.

~3:49 in the full interview

But part of the reason why that happened was because Ghost Rider, the user that did that and hid their identity wisely, just transformed their voice using AI.

~4:28 in the full interview

And so it's trying to find the strengths and weaknesses of that and plugging the greatest weakness which is lyrics really elevates the thing.

~6:48 in the full interview

I think that a lot of the artistic magic is in the crapiness of it, too. I'm not looking for perfect out of these tools.

~12:49 in the full interview

FAQ

What is Burrito Bot?

Burrito Bot is my AI side project that generates songs about Mexican food.

How can AI assist in songwriting?

AI can help in the songwriting process by generating lyrics or musical ideas, but human creativity is essential for enhancing and refining those ideas.

What impact does AI have on the music industry?

AI is reshaping the music industry by providing new tools for music creation, but it also raises questions about authenticity and the role of human musicians.

Transcript

1:24 What's good, everybody, Scoobert Doobert here coming at you, ready to talk about more AImusic and reveal that I have had an AI music project all along, not for very long.I just released my second record in the matter of like six weeks, because I'm really tryingto understand this stuff.I'm trying to make it.

1:43 And a new little bit was discovered and uncovered with Suno.ai and with a new-ish digitalaudio workstation that's able to unlock some of the real cool power of this medium.And it turned it into something that I feel is now, finally, listenable.Is it as good as real people music?Of course not.

2:08 But it's kind of its own different thing, and that's what I'm trying to advocate for inthe generation of this, is not trying to chase just normal music, not trying to put themusicians out of a job kind of thing, trying to make its own weird, bizarre genre.And I think in order to do that and do it in a fun way, it needs to be a parody of itself.That's why I started burrito bot.

2:36 It is my AI side project that only writes songs about really, really good Mexican food.My favorite two songs, if you want to check out this project, are Amole Moon and EnchiladaMama.But the thing that really unlocked it was when I was able to start writing my own lyricsfor it.So if you want an example of kind of how AI does lyrics, I invite you to check out mysub-stack.

3:02 There's a great little write-up, great, there's a little write-up of it, talking about justkind of how AI approaches lyrical structure.And it's ChatGPT, it's lyric studio, they're bad, they're just bad.And if they become a sponsor, maybe I change my mind, I'm just kidding.But they're like, they're like a child, you know, it's not anything beyond what themost basic of basic lyricist would write.

3:35 Because really what it's trying to do is fulfill an expected criteria.You're trying to say, hey, I want you to sound like all of these other songs, and it endsup losing that magic, that playfulness, and it just leans super hard into the cheesiness.You know, certain circumstances, that's fine.Especially if somebody's not trained or whatever or somebody just wants to make a song,yeah, press the generate lyrics button.

3:57 But Suno now allows you to write your own lyrics, and I swear, even with the most minimumamount of human effort, you can make the lyrics so much more fun.And that's what I've been trying to do.And so these those two tunes, I was really proud of a lyric that I wrote.I don't want a sugar daddy, I want an enchilada mama, and that just sparked some joy.And so I set the AI onto that.

4:25 I don't think the AI would have thought of it.This has actually started to come out like across a bunch of different AI generated things,including one of the kind of like breakout AI hits, I guess, actually there too.They were both in the hip hop world.One of them was the AI song, Heart On My Sleeve, that was a fake Drake song.

4:52 And part of the reason why that happened was because Ghostwriter, the user that did thatand hid their identity wisely, just transformed their voice using AI.So the rest was actually made by hand by an actual person.They wrote the lyrics, they wrote the song, they made the beat, and then they transformedtheir voice into something familiar.

5:12 And it's an interesting idea because familiar voices generally lend more credibility andmake things sound better.It's just a weird psychological effect that hits are easier to come by once peopleknow what your voice sounds like.It's proven time and time again.Have you ever noticed that when a singer is successful, their first song is usuallya lot better than their second?

5:38 You don't have to try quite as hard on that sophomore one, it might be difficult, but there'sa lot of examples, I think, out there, and you can just pan through your head.I don't want to drag anybody of artists that are very successful and they release a recordor a single that would have never made them successful in the first place.

5:56 But because they've built a fanbase and because they're a known entity, that thing becomesa hit, despite not really wanting to be a hit.So that's one example.The other example is BBL Drizzy.That is a song that I think was made using Suno that is actually kind of brilliant becausewhat they were doing was they made the AI do a vintage sounding soul vocalist and then

6:23 they put the 808 drums on it and dropped it as a free beat and then so people startedto put bars on it.But the magic of it is the lyrics.The AI didn't write BBL Drizzy and this is a great quote.There is no way AI could write the lyrics like, I'm thicker than a snicker and I gotthe best BBL in history.He adds laughing, classic, awesome, but that's the point, right?

6:48 The AI is able to be like, oh, you want me to transform your voice from this to this?Easy, got you.You want me to take your lyrics and put it into the known entity of soul singer, got you,can do that.And so it's trying to find the strengths and weaknesses of that and plugging the greatestweakness which is lyrics really elevates the thing.

7:10 The other point though is that most of the stuff that's actually doing sort of well likeBBL Drizzy or some of the other things that have gotten sampled or turned into soundcloud beats, they're not just making one generation.People are spending $40, $50 in credits to make a song.I'm not spending that much money because I'm getting creative, right?But there are ways to do this right.

7:42 And the way to do that right is to involve more humanity, not less.And so it's the same thing with AI art, visual art.A lot of that stuff is highly iterative where you tell the AI, hey, I want this.Nope, not quite right.Make it a little more like this.Like this region changed that and it becomes a conversation with the tool.And then at that point, doesn't it become a little bit closer to not AI generated and

8:08 instead AI utilized?Like for example, I just dropped a video on YouTube, if you guys want to check it out,about my new single.And that new single, OMW, has an AI video.And of course, one of the comments and like no shade, I get it.On the other end, I might be feeling the same way if I wasn't using this stuff.One of the comments is like, AI sucks.And I get that.I agree most of the time.

8:40 But what I'm trying to say is I want to see how human artists use the new tools,making things using AI not by AI.And that's where I think the artists need to like plant the flag and tell the story.Because the people that make the tools are going to want to take the credit.When in reality, it's the collaboration, it's the conversation between these tools.And telling it, no, not quite right.

9:09 That's not exactly what I'm looking for.Oh, what if I use this other tool and combine it.And the kind of magic moment for me was when I started to pull the AI generationsinto pro tools.And so now it's not just saying, hey, make a song like this, click straight on to Spotify,throw it in my DistroKid.It is saying, I'm going to generate this 10, 15 times.

9:32 I'm going to try different words, try misspelling, try different genres.I'm going to try and extend it.I'm going to thumbs up, that thumbs down, that do all these different things to try andbreak its mind.That's fake mind.And then I'm going to take all those generations, plug in into pro tools and treat them likeit takes, like you would a musician in the same way that if I hired a guitarist.

9:52 And I said, hey, guitarist, can you play this a little more funk?No, actually, play it a little more doo-wop.Actually, a little more reggae, and then you take all of those and you can kind of combinethem, just like I was talking with vocal, vocal takes, you know, comping.That was the magic that now it's fun.Is it very good, no, but it's fun.And that's kind of the point, right?

10:15 I'm not trying to make a career as burrito bot, but I'm trying to understand the toolsand figure out where I can actually use them, where are they useful?Because what if you didn't do that when the synthesizer came out?What if you didn't do that when the electric guitar came out?Herbie Hancock is herbie Hancock because he used the tools.Obviously, he's a great musician.There are a lot of great musicians, right?

10:42 There's a difference between adapting and trying.And so yeah, I'm not trying to be the biggest AI artist or anything like that, but I'mtrying to make fun of it using its own medium itself.And see, if it leads to something, like best believe I'm going to put a breakdown inone of my own songs using my own lyrics that I wrote, and they have no copyright claimover.I don't want to sugar daddy.I need an enchilada mama.

11:07 We'll put that in a Scoobert Doobert song, best believe that.So manipulating the song in the DAW is really, really cool, and it's taking it to the nextstep with RipX DAW, which is a new one that is purpose built for AI.So I'm still just cracking the services.There's a lot for me to learn, but it is a fun thing when I'm kind of like cooked onmaking other music to just plug away and throw myself a curve ball.

11:38 Every once in a while, I have these little moments where like I wouldn't have thoughtof that.And that's exciting.I think that it would suck though if I gave up like an entire part of my process,like say, for example, like I use rhymezone.com every once in a while.If I can't think of a rhyme.But I'd say like nine times out of 10, I don't actually end up using one of the rhymesthat I see on rhymezone.

12:00 It just kind of jogs my head and I go, oh, yeah, I could think of like a different kindof way to rhyme this.Or I would look at rhymezone.com and see all the different options and be like, hmm, I'llchange the other lyric.I don't want to go there.But it's a cool tool.It's a cool thing.And every once in a while, it's like, oh, I wouldn't have thought to rhyme that vowelwhatever.

12:18 I hope that AI can be kind of the same thing where you have those little moments wherewhere do I go next?How do I switch this up?I'm not feeling this.Let's explore one other opportunity, one other distraction.And then AI can do that for you and then you can move quicker through the process ratherthan replacing the blank sheet.And I think that a lot of the artistic magic is in the crapiness of it, too.

12:48 I'm not looking for perfect out of these tools.I actually have a little frustrated right now by how many of them are trying to giveyou like release ready tracks.I want multi-tracks.I want weird ideas.I want artifacts and decays and genres that have no business being together.I want glitches.I want to feel like I'm looking and poking at a computer's brain.That's cool.That's artistic.

13:18 Because then I can take that and sample it and then play a bunch of stuff around it.Because then instead of like having a laptop on stage, just playing multi-tracks, youhave a laptop on stage and jamming with it in the most bizarre way, in the most non-humanway, in a way that only it could do that thing.It's like I like how synths sound like synths.I don't really like it when synths try to sound like a guitar.

13:49 I like it way more if a synth tries to sound like a guitar and fails.It creates this weird abomination, fake guitar.I like that.I don't know if you can follow what I'm saying, but hopefully I'm at least circling aroundthe bulls eye.I'll just try and make it so you can listen it.Check out the sub-stack if you want to read more about this and hear some of the audioexamples.You know what?

14:13 I should throw on BBL Drizzy on there, too, so you can hear that example.I don't know if that other Drake Hard on my sleeve song exists on the internet anymore.It was pretty incredible.The takedown orders that were rolled out by the music industry, incredible show of forceand power.I'm sure it is on the dark web, but I'm not going to give you instructions on how toget over there.Just use a VPN.Anyway, that's it.

14:43 You had check out Burrito Bot to listen in to my new album, Rap Music, W-R-A-P, musicand check out the blog at lovemusicmore.substack.com and you can hear my music at scoobertdoobert.pizza.Once again, Scoobert Doobert for love music more.See you guys next week.Adios.

The transcript above is a reviewed transcript of this episode (JSON). The recording is the Spotify episode.

Ask a question for a sourced answer, or type a page name to search. Try: