we're experimenting building out a version of this too, but on desktop with www.usebacktrack.com - should have splitting speakers/inputs early next year and seeing what that's like