Legoman
Title of the Song: Legoman
Date of Ideation/Composition: 1/17/25
Background Context: This is my first attempt at making a rap song. Initially, I was just trying to create a backing track with the intent to add additional instrumentation and use Suno to generate a few dozen (or hundred) options to play with later. I had this idea of beat in my head, so as I often do I recorded it via my iPhone’s Voice Recorder app with the intent of either uploading or re-recording the idea later in Suno. Suno has the ability to take raw audio input and output some pretty impressive concepts. In this particular case it was creating a variety of slow jams, chill ambient beats and epic cinematic tracks via variations in the prompting. But one track in particular immediately inspired the idea of attempting a rap song. This article will take you on a journey from conception to execution of a ridiculous rap song about Lego.
How it started
For anyone new to Generative AI (GenAI), prompting is essentially the instructions or recipe you are giving to the tool in order to generate a response. For example, if I want to generate a music track like Mr. Brightside by The Killers (Go Bills!) I would use something like [Genre: Indie Rock, Style: Energetic, Style Tags: Anthemic, Emotional, Beat-driven, Instruments: Electric Guitar, Bass, Percussion, Synthesized Elements] whereas if I wanted to emulate one of my favorite songs La femme d’argent by Air, I would use something like [Genre: Chillout, Style: Dreamy, Style Tags: Atmospheric, Lush, Lounge-inspired, Instruments: Synthesized, Bass-heavy, Electric Keyboard]. There is a bit more that goes into it, but that isn’t the purpose of this article. The most important thing to remember is your output will only be as good as your input, so getting the prompt dialed in is key.
In this case, the initial input was the “mouth music” I mentioned before. I constantly have song ideas pop into my head, so will often try to record an idea of what is going on in my head via what I call “mouth music” or literally just trying to approximate the different instruments, melody, lyrics, etc. as a mixed track by alternating between the different sounds. Recording this does two things for me:
1. If I hear it later on, it will trigger the full song in my head again and I can work on the idea from greater advantage
2. I can upload the idea/concept to a GenAI tool like Suno or Udio and see how it interprets the recorded audio. While generally I have an idea of how I want to prompt it to get closer to the idea I heard in my head, I will often just upload the raw audio, without any prompting, to see what it comes up with initially. As Suno continues to improve, the results have gotten much more impressive. It often only takes a few dozen generations before I have a music track approximating what I’m ultimately trying to achieve. This is what I started with for the uploaded audio on this project:
Another one of my favorite things to do when generating tracks is putting together random styles, genres, instruments, etc. and seeing what comes out. In this case, I wanted something with an epic, dark, intense beat or something along those lines. Well, this is what I heard about a dozen generations later after generating from the track above. The prompt was simply Epic, heavy beat, intense
As soon as I heard it, I abandoned all the other track ideas that had come out by that point (there a few solid generations I plan to revisit later) and focused solely on this one. I wasn’t sure what I was going to do with the track just yet, but I definitely wanted to give it more attention.
To begin, I used the above track as my source track using a feature in Suno called Covers. Covers will take your selected track and reimagine it based on the prompt given. It allows you to transform existing songs by preserving their original melody while adapting them into different styles or genres. This also allows you to experiment with various musical expressions, such as converting a simple song idea recorded with your voice, and turn it into a full instrumental track. This is my go-to feature when trying to approximate the song idea I hear in my head and while it varies in terms of efficiency (a few or few hundred generations), ultimately I usually come away with something that is workable.
Now that I had my source track, I tried a few different prompts to dial it in more. This prompt is what ultimately was giving the most consistency while still getting closer to the sound I was ultimately trying to lock down:
Epic, heavy beat, bass, drum, intense, hard hitting, orchestra
Next I used the remaster feature. This was introduced in Suno V4 and improves audio clarity, sharpens lyrics, and provides more dynamic song structures, resulting in a more polished and professional sound. While I was doing my generations in V4, I have found that remastering will often give a variety of mix options, pulling the highs down or giving a more neutral track. I did this with the track a few times until I had a more even mix to work with going forward.
Once I had a good base track to work from, I started to tweak the prompt to get closer to what I had in mind. In this case, I really wanted the kick drum to hit harder and get a little more intensity with the overall sound. I also added in “orchestra” to the prompt see what if I could get a good string section hit or backing orchestra to kick in with some of the generations. This was the track I chose out of several dozen generations.
While I initially liked how the song just kicked off with a heavy hit to start, I realized it needed a bit of a lead in. So I started generating again trying to get a soft intro, hoping for a piano/keyboard and/or light strings. Less than ten generations later, this came up:
Preparing the Original Song for Suno
Editing and Refinement Process: Now that I had my base tracks, it was time to get to work. First, I needed to pull all the tracks into Logic Pro, a digital audio workstation (DAW) and MIDI sequencer software developed by Apple for macOS, used for recording, editing, mixing, and producing music and audio projects. Once I had the tracks loaded into Logic, I could see where I would need to trim, duplicate, splice, etc. and how I should arrange in the order they should play. Logic also lets me modify the track view to get it down to milliseconds in terms of where to make the cuts. In total, I worked with two tracks, but by the time I was done splitting and splicing, I was up to four before the vocals. Now I had my music set so it was time to figure out how I was going to actually attempt a rap to this thing.
The first thing I needed to do was figure out was the subject. Something I knew well enough to freestyle for a bit to generate some lyric ideas, something ridiculous to make it fun, and ideally something others might find entertaining as well. Looking around my office I saw one of my Lego sets staring right back at me. Literally.
I wanted to challenge myself to have a super speedy “chorus” a la Eminem’s Rap God but also knew there was no way I was going to hit that level of speed or polish. I also needed to figure out the flow, so I just started listening to track over and over and using “bah” and “buh” to serve as placeholders for the cadence and approximate number of words, syllables, etc. I would need to make it match the beat. Once I had a good idea of how everything would fit together, it was time to fill in the actual words.
This next part was by far the most challenging.
How do you go from “bah” and “buh” to actual words that make sense and still fit the theme/subject of the song? This is where OpenAI’s ChatGPT really helped me out. Ideally, I should have started with a prompt that defined the role or expertise I wanted ChatGPT to emulate such as “You are a seasoned songwriter, especially successful at writing lyrics for rap artists for driving rap beats” or something similar, but in this case I used a custom GPT I created specifically for use with Suno called Suno Spark. A custom GPT is a personalized version of OpenAI’s ChatGPT, tailored to perform specific tasks or address particular needs. You can create these without coding skills by providing unique instructions, additional knowledge, and selecting capabilities such as web browsing, image generation, or data analysis. One aspect of the instructions includes this language:
Style Matching: Ensure that the lyrics reflect the style and tone indicated by the music prompt.
Structure: Organize lyrics clearly into verses, choruses, and bridges as needed.
So out of curiousity, I gave it the most basic instructions to see what the response would be:
Output: Here are the lyrics for a rap song about Legos, using terms strictly from the provided reference document:
[Genre: Hip-Hop/Rap, Style: Narrative, Style Tags: Beat-driven, Instruments: Percussion, Synthesized Bass]
Verse 1:
Building up from the bottom, brick by brick,
Colors pop bright, yeah, the flow’s so slick.
Stacking dreams higher, like towers in the sky,
Every piece locks tight, we’re aiming to fly.
Vision in my hands, creativity’s the spark,
From castles to spaceships, lighting up the dark.
Blocks connect, forming shapes so profound,
Lego life vibes, where imagination’s unbound.
Chorus:
Click, clack, pieces snap, hear the rhythm ride,
Legos in my zone, where the limits divide.
Dream it, build it, let the art collide,
In the world of Legos, where we all confide.
Verse 2:
Structure so sturdy, it’s an architect’s dream,
From mini-fig streets to a fantasy theme.
No instructions needed, freestyling the set,
A masterpiece forming, ain’t no regrets.
Modular creations, infinite designs,
Each brick’s a story, each layer defines.
Breaking down walls, remix, redefine,
In this Lego-built world, the power’s mine.
Bridge:
Hands in motion, shaping the impossible,
Dreams take form, every step unstoppable.
Breaking molds, it’s a freestyle parade,
A symphony of blocks, where legends are made.
Chorus:
Click, clack, pieces snap, hear the rhythm ride,
Legos in my zone, where the limits divide.
Dream it, build it, let the art collide,
In the world of Legos, where we all confide.
Let me know if you’d like adjustments or additions to the song!
While this was a good start, it had a long way to go. I began to go verse by verse and query for similar sounding words, but also include prompts like “word like x, that has the same number of syllables as x” or “give me ten different words that mean x, but make sure they are relevant to the lyrics I shared earlier” which really worked well for keeping everything consistent. It wasn’t too long before I had the verses set, but the far more challenging “chorus” took the most amount of time to get right. I had to constantly add and remove words to ensure the flow not only made sense, but was easier to rap as fast as needed to keep up with the music. It was a learning experience, as I found that adding words was actually more helpful than removing words as it let me keep a more natural speaking rhythm and gradually speed up until it was pretty much rolling off the tongue.
After many (many) tries, I finally got the “chorus” down good enough to where Suno *hopefully* would be able to match the lyrics to the melody. Here is the result:
I tried a couple different approaches to mixing the vocals, and ultimately landed on studio reverb with a few tweaks. Ultimately I needed the reverb to cover the tongue-twisting that was happening with some of the words in the super fast chorus. Once I had that dialed in, I mixed the two tracks together and came up with this output to upload to Suno.
Creating the Song with Suno
This is the easiest part of the process unless there is a need to take it back to Logic again and further refine. In this case, since it was more a fun project, I decided to just go with the best output from Suno. At this point, I just needed to pull the lyrics in and generate away using the same prompt. Suno will let you either generate an instrumental track with no lyrics, randomly generate lyrics for you, or let you input your own lyrics. The trick is, you need a vocal track for it to follow, otherwise it will just match the lyrics in a series of random ways. Good if you have no idea what you are going for in terms of the cadence for the lyrics, but bad if you had an idea in mind. This is why it was so important to get the speed of of the chorus down earlier. I needed it to match up with the lyrics in a way that Suno understood so it didn’t try to “improve” the cadence of the lyrics. I was thrilled when after the first generation it seemed to “get” what I was going for, but the voice and mix were still way off the mark.
Luckily, it only took about 20 more generations to get the vocals right and a few more generations for the final track.
Final Version of the Song
Final Track:This is the track I ultimately chose to share, even though I still have a few alternate mixes tucked away that I may revisit someday. The decision to share this particular track came down to a number of factors, but ultimately I really like how the overall mix fits together, especially alongside the intro track, with the kick drum providing the perfect amount of punch without drowning out the other elements.
Initially I toyed with the idea of diving back into the mixing and mastering process to dial in the kick even more, but in the end, the overall feel of this version just felt right. There is always room to tweak things further, but sometimes you just know you’ve hit a sweet spot when every piece aligns in a way that feels authentic and complete. I think this final mix captures that balance and look forward to exploring those other versions down the road when inspiration strikes again.
Conclusion
This was an incredibly fun project! While I’ll be the first to admit my rapping skills are pretty much nonexistant, I’m pretty proud that I could “spit” quickly enough for the AI to handle the rest. In many ways, this project is a testament to creative inspiration: what started as simple “mouth music” meant to back a heavier beat gradually evolved into its own unique focal point.
I’m excited to revisit this piece someday to add more verses, clean up the lo-fi piano, and really refine the overall vibe of the song.
For now, though, this feels like a good stopping point. I’m going to keep exploring the capabilities of Suno and continue honing my recording, mixing, and mastering skills. It is all part of the process, and I hope you enjoyed coming along for the ride!
If you want to make your own music, check out Suno
If you have ChatGPT Plus or Teams, you can check out my custom GPT for creating songs and/or lyrics on Suno: Suno Spark
Finally, be sure to check out my other music, AI and besides, on Suno Soundcloud Bandcamp or Facebook.
j3ff.ai 