Most models accept some form of emphasis notation. It exists because a long prompt dilutes: the twentieth adjective competes with the first, and the thing you actually care about ends up weighted the same as the surface it is standing on.
The discipline is to weight sparingly. One or two emphasised terms per prompt read as intent; six read as noise and often produce a worse image than none.
The better fix is usually a shorter prompt. Weighting is a patch for a prompt that is trying to say too much at once, and shortening it is free.
How many terms should be weighted in one prompt?
One or two. Beyond that the emphasis stops meaning anything, and the output often gets worse than an unweighted version of the same prompt.
Is weighting better than repeating a word?
Yes, where the model supports it. Repetition is a blunt approximation that also lengthens the prompt and dilutes everything else in it.