When creating words I have often run into the problem that too many words sound similar even when I have allowed many different phonemes. Are there methods to avoid too much similarity between words, while making words similar enough to feel part on one language?
A modern approach to this problem is to generate the words automatically according to some formula describing allowed words. You can filter out words that are too close to already generated ones, but usually the randomness used by a computer program will be sufficient.
There are well-known examples of Conlangs with computer generated vocabulary. Loglan and Lojban used some vocabulary lists from natural languages as input for their vocabulary generating algorithm.