Lullabize: Transform Simple Text into Powerful Brain-Building Songs | AI Music Tools
Professor Bear walks through Lullabize, a custom GPT for turning simple text into cognitive-boosting songs, then renders two examples in Suno.
Turning a block of plain text into a singable, memorable song used to require actual musical training. Professor Bear's walkthrough of Lullabize, a custom GPT built for the Lyrical Literacy Project, is a demonstration of skipping that requirement entirely, built on research into what makes simple songs genuinely useful for the brain rather than just pleasant to listen to.
The research behind the tool
Lullabize wasn't built as a novelty. It's grounded in research on best practices for using simple songs in cognitive development, resting on the idea that singing enriches areas of the brain that logic-based learning doesn't reach on its own. That's the premise behind the Lyrical Literacy Project more broadly: getting people, kids especially but older adults too, to engage with simple songs as a form of active brain exercise, since exercising the brain matters more, not less, as people get older.
How the custom GPT works
To make that research usable, the team packaged it into a custom GPT that anyone can open directly. The starting prompt is intentionally simple: it's literally text pulled straight from Mother Goose on Project Gutenberg, along with a small set of commands the GPT will list out for you. The mechanics are equally simple from the user's side: paste in some text, and the GPT applies a set of rules built from that cognitive-development research to turn it into song lyrics, complete with a suggested musical style and meta tags for delivery choices like whispering. Every rule the GPT applies is visible and editable, so anyone who wants to build their own version with adjusted rules, longer or shorter songs, different structural choices, can see exactly what to change. For a personalized song, Professor Bear notes you'd typically fold in details about the actual child or person the song is for: their name, things they like, whether that's Roblox or soccer or something else entirely.
Rendering the lyrics in Suno
Generating lyrics is only half the process; the other half is turning that text into an actual sung recording, and for that the video walks through Suno. For the first example, a reimagined take on "Pat-a-Cake," Professor Bear uploaded his own voice to Suno and used it as the vocal basis for the render. Suno's voice upload feature keeps the output within the uploaded voice's actual range, which matters if the point is to sing along yourself: matching your own natural range makes a song far easier to follow than trying to match an unfamiliar performer's voice.
Two takes, two different tones
The first render came out in a style Professor Bear describes as rough and gruff, which he acknowledges naturally reflects his own voice and taste, but recognizes probably isn't the ideal tone for a song meant for young kids. That realization led directly to a second example: a song about honeybees, deliberately built to be mellower and more chill, still using an uploaded voice but one recorded in a calmer, less aggressive style. The lesson he draws out explicitly is that if you're making a song for your own kid, using your voice or your kid's voice makes it more personal, and it's worth being deliberate about tone rather than defaulting to whatever style comes most naturally to you.
The bigger point: make it personal
Across both examples, the throughline is that Lullabize and Suno together are meant as tools for other people to use themselves, not primarily as a showcase of finished songs from Humanitarians AI. Professor Bear is direct about this: a song is likely to land better with a kid, a friend, or a partner if that person made it, or had it made specifically for them, rather than receiving something generic produced by someone else. The goal of the series is teaching the process itself, prompts included, so viewers can replicate it with their own text and their own voice.
Key takeaways
- Lullabize is a custom GPT built on cognitive-development research into why simple songs help brain development, especially in children and older adults.
- The GPT works from a simple starting prompt, in this demo drawn from Mother Goose text, applying a visible, editable rule set to generate lyrics and a suggested style.
- Suno renders the lyrics into an actual song, and uploading your own voice keeps the result within your natural vocal range for easier singalongs.
- Tone matters: a rough, aggressive style may suit an adult's natural taste but isn't necessarily right for a song aimed at young children.
- The project's real goal is teaching people to build songs themselves, personalized with a name or detail that matters to the listener, rather than only producing finished songs for others to hear.
Try it yourself
Lullabize is available directly through Humanitarians AI's Lyrical Literacy Project for anyone who wants to try turning their own text into a song. Start with a short, simple passage, whether that's a nursery rhyme or something personal, run it through the GPT's rule set, and render the result in Suno using your own voice or a family member's for a song that actually feels made for the person it's meant for.
Full transcript(auto-generated, with timestamps)
[0:00]Okay, Professor Bear here. I'm going to introduce our little custom GPT called Lullabable. Uh, basically what we did is we did a bunch of research on what are the best practices and rules that one would need to use simple songs in cognitive development. There's a lot of evidence that singing enriches areas of the brain that you don't do through logic and other things. It's very, very important for your brain. So the lyrical literacy project is trying to get people to do simple songs, particularly kids and even older people. As you get older, you in particular want to exercise your brain. So to make this simpler, uh what we did
[0:41]Is we created a custom GPT. You can just click on this and go to it directly. I'm going to do it in another window because I just created a little simple prompt to get you started. Um, this little prompt is literally just text directly taken from Mother Goose on uh Gutenberg. So, you can there's a bunch of commands here. It'll list the things you can do. It's very simple. Basically, you just put in some text and all of those rules. All of these rules and you can look at the exact rules that we use down here. This is the exact prompts we're using for the GPT. So if you wanted to create
[1:20]Your own GPT but change the rules a bit, go for it. Uh all the rules are here. You if you don't like this rule, want to add a different rule, want to make them longer, want to make them shorter, whatever you do, you can just change the rules. If you don't want to do that, if you just want to make a simple song, uh I'm just going to do a generic song, but typically if you're doing this for your kid, you'd probably put information about your kid or your friend or your girlfriend or boyfriend, whatever, into the song. like their name, things that they like. They like, you know, Roblox,
[1:53]They like soccer, they like whatever. But I'm not going to do that. I'm just going to take the um Mother Goose text straight up and generate a sum. So it it suggests a style. It has some meta tags about do we want to whisper? Do we want to do this? We want to do that. So um that's it. Basically what I'm going to do for the rest of this video is just play the song that we created. So like, subscribe, do that thing. Uh but the rest of this video is just me going to put that song in Suno, render it, and then play it for
[2:41]The rest of this video. Dor me. Oh, sing with me. M la silly free patty cake bakers woman bake me a cake as fast as you can. Pat it, prick it, mark it with Pete, and hide it warm inside an oven for baby and me. But
[3:53]Wait, she whispers sideways. Enough is never enough. Giraffes glitter fluff, a pinch of moon, a whisper from Mars, candles that sing and smell like stars. She stirs with a spoon of jelly bean wood. dances counterwise as all bakers should. Drops a toad, stool, a tickle, a sneeze, sprinkles of giggle, sprinkles of cheese. As the oven went boom, a flipper slipper ship and suddenly pirates cheering sugar feet. Shiver me treats frosting drips from the sails of the sea. Patty cake. Patty cake. Wildly free. This cake is a spell for you and for me. If it wiggles or barks or laughs like a bee, feed it a
[4:59]Rhyme and a sip of iced tea. Oh, feed it a rhyme and a sip of icy. Okay, Professor Bear here again. I realized when I made that video, the one the patty cake baker bakers man, I use my own voice insu. So I have a grabbly nasty voice. So I upload it to Sunno and use that as a basis. You'd basically if you go to Sunno here, I think we are in Sunno here and you go to create. I don't know why this isn't you could just upload audio and upload your own voice. So that's what I did. I uploaded uh my voice to uh
[5:51]This patty cake salon, but I do it in a style which is sort of very rough and gruff and realize that probably isn't the best for kids. It's great for youth. So, if you do this, it's probably a good idea to upload your kid's voice or your voice because what it does is it improves your voice, makes it a better singer, that kind of thing. But it keeps it in your vocal range. So if you want to sing along by using your voice, it's closer to your voice. You know, if I'm trying to sing Whitney Houston, good luck. But if the AI created thing is sort of already close
[6:30]To what you are, it's a lot easier to sing along and follow along. And that's sort of the point of the lyrical literacy of point of the lullabal eye stuff is to actually do some singing and to have this with your kids or yourself or whatever. Um, but I realized when I did that, I just sort of naturally did it in a style that I like. I don't like a purely AI voice. I like the fact that my voice makes the AI sort of grittier and worse, not as perfect to me. That sounds better. Uh, but when I did that and listened to the song, I
[7:07]Realized that if you're making it for your kid, you probably want it for, first of all, in their voice or in your voice, so it's more personal. So, I'm going to redo the song. I'm just going to make another song about honeybees. And purposely, uh, it's still going to be my voice, which I use, but I'm going to use a voice which I uploaded where I spoke less aggressively and less, you know, just more chill. Uh, so I'm just going to say make a song about honeybees. Need to spell these right. And then it's going to go ahead and do that. It's going to start doing that.
[7:46]And again, I'll render it and play it. This is going to be a much more general lai type thing that I'll intentionally do. I won't do what I did here, which is what I do naturally, which is sort of make it rough and aggressive because I just personally like that sound. Uh but if you're going to do this uh like for your kid in particular, you probably want it sweeter in general. It's not this which is making it. It's the way we deal with suno which gives it that sound. We'll have a lot of other videos on how to get whatever sound you want out of syno including with your own
[8:18]Voice. But again, I after I listened to that song, I realized I just sort of naturally automatically without thinking it did it in the style that I normally do stuff. And that's probably not the best style for like Don Laai for a kid. So, we'll see how this one turns out. Honeybee song. Um, but I'm I'm going to intentionally think about it making it mellower and chiller. Particularly with the text. I also do things here where, you know, I edited a little bit. You know, I added little annotations. I changed the name. It's a Patty Cake, Baker's Woman, because Patty Cake just sounded like a woman's name. So, I made
[8:56]Some changes. But for the Howdy B song, I'm going to intentionally keep it like I don't get for a kid. Swarm of bees in May, buzzing bright through the day. Fills the fields with golden bloom. Treasure worth a haze stacks room. A swarm of bees in June sings a sweeter summer tune. Honey's gold, they were divine. A bounty like a silver shine.
[10:14]A swarm of bees in July. When the flowers wilt inside, they hum a song that fades away for their worth has gone array. Yet bees obe all through the year. Their work brings life, their hum brings cheer. The July bus holds little weight. They labor still we celebrate. From clover fields to fruit trees, the world survives because of bees. So May to July, their worth may swing, but to life itself, they're everything.
[11:46]Okay. So, I think that second song came out much mellower and more like uh something you'd make for your kid. I going to end this here. I'll do other videos on like here's the took me about an hour or two to create both of those uh the footage for both of those videos, the Patty cake and the Bumblebee one. Uh we won't get into that in this. Uh but my main thing for all these series is to teach you how to do it. Uh, I've had people who aggressively get nasty and hate my voice or hate my art. Whether you like it or don't like
[12:26]It, I don't really know. But I do know that it's probably better if you make it for your kid or your best friend or your girlfriend or boyfriend or father or mother, whoever than if I make it or Humanitarians AI or Bear Browning Company makes it for for you. So, uh, Vom stuff, uh, you know, we have a Spotify, we have stuff, you know, stuff on the YouTube now which you can listen to, but I think most people are going to rather do it themselves. So, a lot of the focus of this is going to be um, how to, you know, create little videos like this, how to create the art
[13:07]Is done with a pecode, by the way, but a bunch of PCODE stuff. This it's also done with the PECODE for the Patty Baker thing. Uh, normally I would spend a lot more time on the art getting the style right, but because I wanted to get this video out quickly, I, you know, did both of the videos in probably total of two hours, including generating the art styles, but we'll talk about in other videos. So, uh, get do the like, subscribe thing. I'll see you later.
More videos
2:08Bridging the Pixel Gap in Browser Automation.
2:23How One Narrow Safety Rule Can Make an AI Less Safe Everywhere Else.
2:04Why splitting a chunk from its document makes it retrieve for the wrong question
4:20Three You Can Take Back. One You Can't.
2:21Why a 50-turn agent pays for the same screenshot 35 times unless it caches the pixels
1:53