In creating a set of sounds for a language, there are many possible approaches. The one my friends picked, when we first set out to create a language 5 or 6 years ago, was the method of picking sounds we thought were cool, and lumping them together in a language. The problem with this method is that the way in which sounds are strung together is somewhat formulaic. In part, it depends on what sounds can be made by the species on question. As I have said, Linguistics and Fantasy alike ask ‘what if’. Phonetics is no exception. What would a language look like, were it spoken by beings utter incomprehensible to us?
And this creates the Cthulhu factor. This is demonstrated most clearly by H.P. Lovecraft, who deliberately picked unpronouncable sounds, stringing together awkward combinations of consonants, and declaring them only close approximants of the actual sounds. The implication is then that the creatures in question are so alien that even in a loosely translated form, it is basically impossible for humans to understand. This is all good and well for Lovecraft, who had a universe of incomprehensible and wholly alien beings. But what about for those of us who want to create languages for human or near-human beings (or of any specific type of being, for that matter)?
To pick sounds the human brain can hear is not so simple as taking all of the sounds present in English, however, and giving them all their own letter. This would create a language with a better letter-sound ratio than English, but it would not cover the array of possible sounds. Some African languages involve clicks, Swedish and Finnish have a few vowel distinctions that aren’t made in English, and so on… So we need a method of showing all possible human sounds.