You need not limit yourself to these common stereotypes; however, I would not advise you to write every character's dialoge in their language. Readers hate having to refer to the back of the book every time someone speaks. It slows the story down, and quickly frustrates the casual reader. If you're so desperate to brag about your language-making skills, write a dictionary or something. But don't bore the readers. If you're looking at writing as a profession, they're the ones who are paying for your meals. You would do well not to displease them.

The best way to make a language is to base it on one or more languages that already exist. It doesn't have to sound anything like any other language on Earth, but it should at least follow some of the same patterns. For example, in Spanish, "I do" "You do" and "We do" use completely different words for each "do". In Japanese, you can completely change the context of a sentence by adding or taking away suffix phrases, such as "wa" and "no".

These are some of my attempts to make a language and what went wrong:

First: Old Kiish

This language was completely disorganized. I picked random words which sounded cool but didn't necessarily have any root or significant similarity to other words. I didn't set up a system of grammar, using English grammar and substituting words when I couldn't find them on my lists. I often made two, three, and sometimes four words with exactly the same sound and spelling--but entirely different meanings--without realizing it until afterwards. All in all, this language was very pretty but very impractical and extremely lacking in organization.

(Blah blah blah blah... Right? You know, you could probably write "Blah" into your language so that it actually means something.)

Second: "Modern" Kiish

This language tended toward the oppoosite extreme from the first: too much organization. Every word was formed from a group of smaller, less specific words. However, this could lead to problems. Base words included "water", "move", "big", "mind", and other general terms. Things like "think" and "person" and "ocean" were built from lots of little root words. Lots and LOTS of little unaltered root words. So many in fact that one began to lose track of the original meaning. For example, "human" was a combination of two words, "learn" and "animal". However, "learn", was also made up of multiple words: "think" and "keep". "Think" comes from "mind" and "move". "Keep" is a root word. "Animal" comes from "life" and "move", the latter being a root word. "Life" comes from "Water" and "Magic", the latter coming from "Mind" and "Create". So, in the end, the word for "human" ended up being "Mind-move-keep-water-mind-create-move". Theoretically, this string of seemingly unrelated ideas would make the speaker of kiish go "Ah! You're talking about a human!", but many fairly basic words such as "human" ended up being a considerable mouthful, and actual conversation in this language would be so slow as to be unbearable! I decided to abandon this simple yet alarmingly complex language in favor of saving my poor tongue the effort. At this point I still had no effective system of grammar; everything still sounded remarkably like English despite small differences in pronounciation.

Third: Shentaro

This is so far the largest language I have constructed. Of all the languages, Shentaro is the one that most resembles English in design. There are rules, but they apply to specific groups of words that may have nothing to do with each other, rather than the whole language. Everything is exceptions. Some of the words resemble English or another language. For example, "mesmara" means "beautiful", and it sounds like "mesmerize". However, other words, such as "yamanik" ("hope"), are entirely unique and have nothing to do with any other language. I didn't come up with a grammar system for this language until after I had already completed 23 pages of words, and now I deeply regret it. This was a wonderful language, and I even learned to speak it relatively well, but I just couldn't figure out how to put everything together without destroying parts of it.

Fourth: Touthan

This is the language I am currently working on, and so far it's doing pretty good. The grammar system mostly resembles Spanish, although there is no "the" to be found. Each verb specifies tense and subject. In other words, the verb "to be" has many different forms: I am, I was, I will be; You are, You were, You will be; He/She/It is, He/She/It was, He/She/It will be; They are, They were, They will be; We are, We were, We will be. Certain rules apply to each of these forms; therefore, you only have to know the "to" form--"to be"--to know what each form should be. There are no exceptions to this rule. Also, all of the adjectives end in "r" or "ar", and all of the "to" forms of verbs end in "sa", so many verbs and adjectives are actually nouns with "ar" or "sa" tagged onto the end.