I'm interested in sparse, interpretable and easy-to-manipulate representations of audio.
https://blog.cochlea.xyz/sparse-interpretable-audio-codec-paper.html