I took a research track with my Double Majors in Comp Sci and Music education. My senior thesis is an application or A.I capable of transcribing any song into sheet music. Motivated by my frustrations in finding free and reliable transcriptions for African American music, from gospel to hip-hop etcetera. But also guided under my deep passions and understanding of both Mathematical Algorithms and Musical instrumentation.
I consider this my magnum opus. My Bankai (卍解).
Once ready, it will be released to this webpage 😊.
Enjoy the evolution of Zion.
Sophomore year of high school, I wrote a program on my Texas Instrument Calculator that did music theory homework for me. The animations that involve the movement of textual characters forming the shape of a treble clef are coded through mathematical equations. While technically it doesn’t transcribe music per-say, looking back, I still felt it was pretty brilliant.
Like, awwww look a me. A lame little nerd. lol.
Zion: Iteration 2 (2022)
I wrote this one sophomore year of college (UWM 2021) on the walls of a psychiatric hospital. I had just been drugged and sexually assaulted, resulting in a manic episode where I broke all of my musical instruments and threatened my own suicide over all social media. 4 weeks upon release I took all of that code, translated it to Swift (Apple’s bootleg version of Java) and I shoved it into an 2nd Generation iPhone SE. Somehow,,, it worked lol
But in retrospect and given the context of today’s technological advancements, it is merely a guitar tuning app that I nigga-rigged into a note detection and placement program.
Baby steps, I guess.
Zion: Iteration 3 (2025)
Perhaps this is where things get interesting. This is A.I…
As in like, the code is a python script, an small, trained, audio model. (Perhaps that doesn’t sound as cool as a Large Language Model). Um. Yeah. It took me years to grasp the fundamentals of modern Machine Learning Algorithms. My Internship at Georgia Institute of Technology helped tremendously. Improving upon iteration 2, this one has the capability of detecting rhythms, even for melodies with harmonic layers. Iteration 2 could not detect more than one note at a time.
I trained this algorithm on my own transcriptions. Having transcribed for multiple HBCU bands and recorder covers for TikTok clout, I had a plethora of reliable training data.
Zion: Iteration 4
Coming soon…
