I'm working on a project right now that's essentially a rhythm-game (think Guitar Hero or Dance Dance Revolution) style app. It needs to listen to input from an instrument playing notes on a scrolling score of music.

For the pitch tracking I only need to track monophonic voices, so this isn't a big issue and there are a lot of options— however I'm having a bit of trouble with the rhythm tracking of each note. I need to detect precise attack times for each note, at least at sixteenth note accuracy, hopefully better.

So far I've been using sigmund~'s note mode to get the pitch and attack at around the same time, but there seems to be a bit of a delay at any combination of npts and hops sizes. I've also tried fiddle~ for the attack detection, with sigmund~ taking care of the pitches.

Bonk~ seems to be interesting for responsive attack times, but there's the possibility of non-pitch noises being picked up as well that are not related to the instrument.

I think it's just going to take a lot more experimentation on my end, but would anyone have any advice on a good strategy for accurately tracking the articulation and pitch of notes in real-time?