Oh, Cool!!
Thank you!!!
Let me ask you a question, did you have a chance to check the converter I uploaded? by chance, do you find it at least decent? does it work differently than the one you uploaded?
I ask all of this because, as you may have read, I am not a programmer and used AI to help me iterate the versions. At this point, I honestly do not know whether I should continue iterating or just wait for people like you (that most likely actually know what they are doing) to improve on it.
BTW, if you are the person that created degausser, THANK YOU SO MUCH!!! I was able to load some new (bbp) songs into my copy of Daigasso.
Thank you (and everyone) for the amazing work.
Hi, yes I'm the creator of Degausser (from the old DX Degausser to the more modern Degausser for bbp, and also degausser3ds), though I've been away from the scene for a while, so it's exciting to see things are still happening.
I did have a look at the converter, it's certainly a step up from the original bbp2midi.py which crashes on a lot of stuff. But fundamentally I don't think either the piano chords or the guitar chords are right. And there's no way for the AI to independently rediscover this either since it's really just reverse engineering my Degausser work which already didn't play these correctly.
Roughly speaking, there are two flavours of piano chords (1. generate from root+accidental+stuff, 2. four literal notes), and yours gets it wrong on both counts. So the piano chords is basically random noise.
And there are three flavours of guitar chords (1. generate from root+stuff, 2. six literal notes, 3. bdx migrated), Your AI-improvement gets the third one right, and about half of the "six literal notes" version. So you can actually get decent guitar behaviour. Though it still treats consecutive strums as a legato, and also ignores strumming a subset of strings etc.
The other thing which doesn't behave the same as volume control, and I'm not convinced I have the right formula either. Because there's the master volume, per-instrument volume, per-track volume changes, per-note intensity changes etc., and they combine in mysterious ways. I've made my best guesses which sounds reasonable, but not 100%.
But keep it up anyway! It's exciting to see what AI (even at the free tier) is able to achieve. Maybe even feed my website back into it and see whether it thinks my version is better, could be a fun experiment.