Lyrical Literacy GPT Tutorial: Create Educational Songs in Seconds! | Humanitarians AI
The Lyrical Literacy GPT turns simple text commands into ready-to-sing lyrics loaded with meta tags, built to help anyone write personalized songs fast.
Writing song lyrics from scratch, especially lyrics loaded with the specific meta tags that make Suno or Udio productions sound professional, is its own skill that most people don't have time to develop. The Lyrical Literacy GPT exists to skip that barrier entirely, turning a short command and a bit of source text into a ready-to-use, fully tagged song.
A command-driven tool
The tool works through a straightforward command list, and typing just "command" into the GPT shows the full set of available options. Behind that simple interface sits a large body of underlying data, including a substantial library of poem structures and metatags. That metatag library alone spans hundreds, possibly thousands of entries, covering musical genres like African blues, blues rock, and British blues, along with a wide range of instrument choices, vocal techniques, and other musical parameters, all of which the tool can pull from automatically depending on the command used.
Turning a fairy tale into a children's song
The tutorial demonstrates the tool by generating a children's duet using the "child duet" command, paired with source text pulled directly from the Grimm's fairy tale "Hans in Luck." The "child" portion of the command instructs the tool to keep language simple and easy to sing, appropriate for a young audience, while "duet" sets up the song for two alternating voices. Pasting in the fairy tale text and running the command produces a finished, singable song built from that source material, turning a piece of classic literature into something a kid could actually sing along to.
Why the underlying mission is about singing itself
The point of the tool isn't really the technology, it's getting people to actually sing, because singing functions as a form of cognitive exercise. The brain, like any muscle, can decay without regular use, and while activities like logic puzzles train specific cognitive skills, they don't train much else. Singing is different: it activates many regions of the brain simultaneously, which is why the Lyrical Literacy project treats music as a broad, low-effort way to support brain health, whether the goal is helping an older adult guard against cognitive decline or helping a young child learn to read.
Iterating quickly
As with most generative AI tools, the first output from the Lyrical Literacy GPT is rarely the final version. The natural workflow is to generate a song, evaluate what works and what doesn't, and then iterate: maybe one section needs to be richer, maybe the style needs to shift, maybe specific lines need reworking. What makes AI tools valuable here isn't getting a perfect result on the first try, it's how quickly a person can cycle through revisions until the song actually feels right.
From lyrics to a finished track
Once lyrics are generated, the natural next step is bringing them into a platform like Suno to actually produce the audio. That means pasting in the generated style description, choosing a title for organizational purposes, since a clear title makes it easier to keep multiple in-progress songs straight, and then dropping in the lyrics themselves, which already carry the relevant metatags baked in, things like a warm-up section, alternating female and male voices, a specified raspy quality for the male vocal, and instrumentation choices like banjo with a twang.
Personalizing the voice itself
Beyond just the lyrics and style, tools like Suno and Udio also allow a user to upload their own voice recording. The match isn't perfect, but the tool will generally keep the output close to that uploaded voice's range, meaning someone can create a personalized singalong track that actually sits comfortably in their own vocal range rather than a generic AI voice, and the fine-tuning process to get there tends to make an uploaded voice sound noticeably more polished than an average untrained singer.
Key takeaways
- The Lyrical Literacy GPT works through simple text commands, with a full command list available by typing "command," backed by a large library of poem structures and musical metatags.
- The "child duet" command demonstrated in this tutorial keeps language simple for young audiences while setting up alternating vocal parts, illustrated using text from the Grimm's fairy tale "Hans in Luck."
- The broader goal behind the tool is encouraging actual singing, since singing engages far more regions of the brain than more narrow cognitive activities like logic puzzles.
- Iteration is expected and encouraged: generate a first version, evaluate it, and refine the style or specific lines until the song feels right.
- Uploading a personal voice recording to a platform like Suno or Udio lets the generated song stay closer to a person's own vocal range rather than sounding generic.
Who this is for
This tutorial is aimed at teachers, parents, and anyone who wants to create personalized educational or singalong songs without writing lyrics from scratch. It's part of Humanitarians AI's Lyrical Literacy program, which is built around the idea that singing is a practical, accessible way to support brain health and learning at any age.
Full transcript(auto-generated, with timestamps)
[0:01]Okay, Professor Bear here. Uh, I'm gonna tell people how to use our little lyrical literacy plugin. Uh, it's a little tool that helps you write songs for Suna or Udo or whatever for your kids, for your parents, for your friends, for whomever. So we have a command list here and this has different kinds of uh factors just to let you know if we want to look at how this thing was built. We have a lot of data that we're using and particularly around some poem data and some metatag data and what I mean by metatag data is this uh right here. Uh so these are metatags.
[0:51]So we have hundreds maybe thousands of metatags you know African blues, blues rock, British blues, medium blues, different kinds of instruments, different kinds of vocal techniques, different kind of musical techniques, etc., etc. Uh and so what this little tool here does, it has a bunch of commands. If you just type in the command command, it will show you its commands. But if for example I want to write a children's song which is a Jew duet I just put in child duet and then I'm going to do this because it's a kids thing for just a fairy tale. So this is the actual text from a Grim's
[1:31]Fairy tale called Hands in Luck. So we're going to make a little Hands in Luck song here which is a children's song. What that's telling it to do is keep the language simple and easy to sing. So we don't want super complicated stuff. We just want, you know, the point is the lyrical literacy product is project is to get people really actually singing because that's basically exercise for your brain. Uh whether you know it or not, your brain will decay like your muscles. So if you do logic puzzles and whle that's great, but that only trains logic. It doesn't train many other parts of your brain. So just doing
[2:08]Rert is kind of like working your biceps but no other aspect of your body. The point of the lyrical literacy project is that music actually involves many many regions of your brain. So whether you're young or old, if you want to prevent Alzheimer's or you just want to learn to read better as a little kid or want to learn to read better as an adult, uh you need to get your strong brain. You get and music is a great way of doing that. And the way you do it is to do it. So create little personalized songs and this tool is to help you do
[2:39]That. So I just go child duet. I paste in my thing here and it gives me a song. As with everything in AI, there's a lot of iterative changes. So typically what you do is you generate a song and go, well, I like this part of the song. I don't like that part of the song. This needs to be richer. Maybe I change the style. Maybe I change this. But you get started. What's great about AI is you can very quickly ideulate very quickly iterate. So what we're going to do is we're this is just random suggestions. You can say I want an R&B song and it
[3:16]Will make an R&B song. But we just ask for a song. So it's just going to randomly pick from the bazillion tags we give it. And then we go to our suna and we put in our style here. And then we uh put in our little title here. Again, the title can be changed. Uh, but it's nice to have a title. That way, you can tell it from your other songs that you're working on. So, we're going to give it a title. And then we put it in there. I'll play these songs at the end after I uh make them. I'll display them at the
[3:53]End of this video after I finish speaking. Uh, and then uh we just put in the lyrics. And what I typically do, and notice, let's look at these lyrics a little bit. This is where the metatag the metatag is saying doing a warm-up, a little, you know, warm-up thing. First a female voice and then a male voice and the male voice will be raspy and there should be a banga banjo and twang and yada yada yada. So what I do is I just hit a few. I'll play these at the end. Uh a few samples of the end of this video and that's it. So again, the link
[4:24]To the the tool here. So, this is the the long link. That's not a very convenient link. So, we'll put that link in the YouTube description, but our lyrical literacy tool just helps you begin to make your own songs, whether it's for yourself, whether it's for, you know, your kids, whether it's for your parents, whether it's for your grandparents, whatever. Uh, what's great about tools like these, it's very, very quick to make little songs. You can even upload your own voice to things like Suno and Uo and it's doesn't perfectly match your voice, but it'll keep something close to your voice. So, it'll that typically means
[5:01]It'll keep in your vocal range. So, if you want to make a little singalong song in your vocal range, you can just upload a high quality recording of your voice to Suno or Udo or the bazillion other ones that are coming out and it'll do something pretty nice, pretty good. It'll actually typically make you a better singer than you are. Uh because that's what it does. It just, you know, it takes the average of the best singers and integrates them. But if you put your own voice, if you upload your own voice, it'll actually try to an extent match that voice. Okay, that's it. I want to
[5:34]Keep these videos super short, super quick. Again, like, subscribe, ring the bell, whatever. And I'll do another one soon. Don't rain me. So lucky. Can't you see? I worked for seven faithful years. A silver chunk I earned with cheers. But all this weight me so lucky. Can't you see? I work for seven faithful years. A
[6:47]Silver chunk. >> Don't rain me. So lucky. Can't you see I work for seven faithful
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