Utilizing AI in Music Creation
It has been a few months building Bovo Music, a music platform that leverages next-generation AI foundation models to understand music theory at its core. So what we are working on is very simple,
We want to help user transforms melodies into fully arranged songs while providing unprecedented creative control over the final result.
What we do
What we do starts with something very straightforward: we take a melody input and generate a lead sheet.
While this may seem like a simple task for professional musicians, it’s not as obvious for those without sufficient training in music theory.
The reason is that harmonizing a melody is not just a mathematical calculation — it’s a two-dimensional process.
Imagine a middle C in the melody. One might harmonize it with a C major chord, but it could also be harmonized with Am, Ab, F#dim7, F#m7b5, F, Eaug, Dm7, Dm7b5, Dbmaj7… the possibilities are endless. The choice also depends on what comes before and what follows. It’s the progression that makes it complex.
Bovo Music takes care of it.
What we did after talking to people
While the lead sheet excites me, it doesn’t evoke the same kind of excitement — or goosebumps — for many others. The reason, though not obvious to me at first, became clear: the only difference between a melody and a lead sheet is the addition of chord symbols above the score, and for most people, that distinction isn’t immediately meaningful.
Our team quickly recognized this issue and developed an audio playback function that supports multiple accompaniment patterns on the piano. This marked the beginning of our transformation — from symbolic notation to actual music.
We also allow users to adjust the ‘wildness’ of the harmonic progression, giving our foundation model more freedom to generate harmonies.
Our first beta user
We started beta testing in March 2025 (literally just a few days ago), and we’re already amazed by what our users can do.
One of our beta users is an indie game developer who prefers creating his own music rather than using stock tracks. While he previously trained as a pianist, he lacked equivalent training in composition — something that would have helped him complete the task.
Using Bovo Music, he started with his own melody, generated the harmony multiple times until he found one he liked, and then simply exported and arranged it in a DAW (Digital Audio Workstation).
For privacy and copyright issue, I have replicated the process below rather than using his music.
After a few rounds of chord generation, this is the one I liked the most.
So, I downloaded the MXL file and imported it into Logic Pro. Similar to what our beta user did, selecting the right sounds in Logic Pro was an enjoyable task. He showed me his selections and demonstrated how he added additional audio filters to make them sound “cool.”
I tried my own version, with different VSTs, and my own taste of audio processing.
Here is the final results.
If you want to be a part of the Beta Testing Group, reach us at hello@bovo.ai.
