Linux

Here's an XKCD describing the development situation you just described.

Converting audio into a different set of tunes, especially if they have multiple layers, is ridiculously difficult. There already exists MIDI converters online a google away that can create chiptune-esque arrangements from input songs but they SUCK, doing nothing more than trying to match input frequency averages to the target frequency of every midi note. But just asking someone to create an app (let alone a free open source one) that could make a usable chip tune with the slightest awareness of its input tunes is a very very tall order and this is not the right community to be asking it.

So I work in game audio. I have a masters in music composition, and beginner knowledge of coding just to add a few lines of audio-related code to my client's spaghetti code.

What you're asking for is not a simple task, and will certainly need AI or machine learning or something along those lines. You could more easily make something like this if its just a single melody with no other backing tracks (and there's plenty of software that does this), but to understand and distinguish separate instruments, including drums, hear them all as separate voices, and turn them into a whole other medium, is on a whole other stratosphere. If someone's going to make something like this, it's going to be made by a corporation and sold as a $1000 plugin.

But hey, at least it will most certainly be in C++, because that's the language of choice for most audio plugins.