Characterising continuous effects in linear mixed models as fixed or random


I've started to learn about linear mixed models, and I'm a bit unclear on the distinction between fixed versus random effects when the effect is continuous. Let's use an example where participants react to pictures (dependent variable: reaction time) and these pictures are described by arousal and contrast which are both continuous predictors.

I've read that random effects are those that you want to generalise to other types of the same thing, so participants and pictures would be a random effect. However, if we have continuous predictors describing these pictures and we'd like to investigate their influence on reaction time, then we will probably never have all values of the continuous predictors. So, would arousal and contrast be fixed because they are of interest but random because we'd like to generalise?

Or does this mean that we have arousal and contrast as continuous fixed effects and picture as a random effect? If yes, doesn't the model try to estimate how much variance is explained by the random effect, and wouldn't that "take away" from the effects of interest?

Hopefully, I have described my thoughts understandable enough, I'd love to understand this better.