...Oh, more specifically, I have notes on the english language that include most of the consonant clusters (Too many end clusters lol). However, since Arpasing is diphone-based, I broke down everything into 2 phoneme units (ex 'str' -> 'st' and 'tr') and deleted the duplicates. This left me with the raw diphones I needed for the list.
However, you may still want to record more complex clusters anyway, since (for example) a phoneme like "tr" could sound different depending on whether it came from "str" or if it's by itself. Context-based phoneme selection by Arpasing Assistant will account for it.