Calls that use EnvKey.track_voice to infer things about the track. For example, you can automatically have the first and second tracks get hand=l and hand=r, respectively. If you use the local definitions file to bind >pno = infer-hands, then every >pno track will both set the instruments and the hand. Only explicitly named instruments are counted, so you can still have multiple tracks for one hand, if they are named >.

TODO this is experimental, because I'm not sure I like it. It means parts can change just by inserting a new note track, which seems non-obvious.