Does load_confound unit the confound variance?

Hello everyone!

I have a quick question on Nilearn’s load_confound: regarding this documentation, load_confound does demean the confounds but it is not clear whether it units the varaiance as well or not? Since it is suggetsed to z-score the confounds (power et al., 2014), should I turn off the demeaning in load_confound and then use standardize_confounds=‘zscore’ in masker.fit_transform?

Many thanks,