Linguists have lists of characteristics of (human) languages called linguistic universals. They include complex (not the same as complicated, for those who don't know) syntactic rules with complex interactive properties, complex lexical items (made up of syntactic, phonological, and semantic features), complex phonological rules that interact in interesting ways, and semantic principles that account in part for how we understand more than is actually said or written. Linguists have guided the creation and evolution of some of these popular-media communication systems, but none of the systems has approached qualification as a language in the linguistic-science sense.
Yes, there have to be people who speak it, but evolution is involved there, too. When people who speak different languages try to communicate, they evolve a system called pidgin, made up of syntactic, phonological, and lexical elements from their native languages. Pidgins evolve and become enriched over time, with use. Sometimes pidgins die out and are replaced by other languages. But sometimes a pidgin becomes so enriched that it approximates the full set of linguistic universals, and becomes the primary language of a population. When a pidgin is so established and its use so habitual that it becomes the native (first) language of a generation, it is called a creole. The native speakers of creoles impose still more human-language characteristics on their languages, and these then become full languages in their own right.
BTW, the size of the vocabulary has very little to do with qualification as a real language, but it's the aspect most fans of popular-media communication systems pay attention to.
And for anyone who objects that the languages of non-humans shouldn't be subjected to the criteria that human languages have to meet, well, I agree, but then the question of whether they're "real languages" quickly dissolves, doesn't it?
DA Wilson, Retired professor of linguistics (and HTG fan)