when i was 16, i made a crappy app to continue midi music. gradually over time i gained experience and it became better and better, until it reached, at times, imperceptible to real, human midis.

kobimusic is no longer that app; it's a collection of ai-music projects of mine with the goal being to 'empower the musicians' (i know, vague corporate mission; bear with me).

first, before i talk about the 'mission', i need to address something. to be blunt, a lot of ai-music related apps are what i like to call 'ChatGPT-wrappers'. they strap on an oai api onto a daw, and call it an 'ai powered daw' for instance.

there are some cool companies that don't do that (shout-out to friends at songscription, staccato). but most do. and honestly, in my time in this space, not really too much has changed; most of the 'best' work is 3+ years ago.

anyways, back to the mission. first and foremost, i want kobimusic to be an oracle of real, sota, research that people actually use. no chatgpt wrappers here. secondly, i need to elaborate on what the mission means. i want to get rid of friction from music.

if you're a pianist, you want to play your favorite songs. arranger, you want to re-imagine your favorite songs. historian, preserve music. composer, make new music. my point is many people want to do many different things with music, but not every person enjoys every part of that creation process. with time (and gpu hours), i'd like to address each one's needs here. i'll get more into how with the planned projects section.

planned projects (probably around this order)

1. omr model / engine
2. midi / audio alignment algorithm
3. music transcription models
4. ai instrument vst
5. —
6. also general small projects to be released