When recording tracks from DAWs, we could potentially optimize how notes are assigned to voices. This mapping affects the entropy in the resulting binary (because there is ambiguity that affects the byte order in the binary without changing the resulting song, that could be used as a mechanism for optimizing binary size).

For this, we descussed briefly that a systematic investigation could be made to get insights on how the current strategy performs for size-coding and whether or not there are any strategies that could improve binary sizes through smart voice assignment.