So this is actually more tricky than I originally thought, since the bigram creation widget only works for transcription symbols, not orthographic. To make it work, it would have to detect whether spelling or transcription is selection, show the correct inventory, and also have checks that the bigrams created are actually relevant for spelling vs transcription. Since it's a lower priority thing, given that no one's used MI for orthography in linguistics (to my knowledge), I reverted the change to just use transcriptions and tiers.
So this is actually more tricky than I originally thought, since the bigram creation widget only works for transcription symbols, not orthographic. To make it work, it would have to detect whether spelling or transcription is selection, show the correct inventory, and also have checks that the bigrams created are actually relevant for spelling vs transcription. Since it's a lower priority thing, given that no one's used MI for orthography in linguistics (to my knowledge), I reverted the change to just use transcriptions and tiers.