As I recall from the recent stream, different colours also mean more polygons in-game; on the other hand, most (all?) default SH models are not hollow, so you’ll probably be okay there if you don’t bother hollowing the models out. Also of course, every edge of a voxel is more polygons to render, so try for straight lines rather than curves where possible.
What else… Stonehearth models tend to have little shading in the voxels themselves. On the other hand, if these are rare beasties (ie more than 1 on the screen at once is exceptional) then you should have more freedom to make the models more complex, because nobody’s PC will have to slog to render loads of them.