support for combining two samples of the same talker in each input item