A brief is not a wish. It is a shape imposed on a space of possibilities, and its geometry decides what can come back.
Constrain every axis and you get exactly what you already imagined, which is rarely worth the trouble. Constrain nothing and you get the statistical centre of the training distribution, which is worse. The useful briefs are tight on the axes that carry the meaning and deliberately loose on the axes where you want to be surprised.
Knowing which axes those are, for a given piece of work, is most of the skill.