ABSTRACT We present simple xspec models for fitting excess variance spectra of active galactic nuclei. Using a simple Monte Carlo approach, we simulate a range of spectra corresponding to physical parameters varying, then calculate the resulting variance spectra. Starting from a variable power law, we build up a set of models corresponding to the different physical processes that can affect the final excess variance spectrum. We show that the complex excess variance spectrum of IRAS 13224−3809 can be well described by such an intrinsic variability model, where the power-law variability is damped by relativistic reflection and enhanced by an ultra-fast outflow. The reflection flux is correlated with that of the power law, but not perfectly. We argue that this correlation is stronger at high frequencies, where reverberation lags are detected, while excess variance spectra are typically dominated by low-frequency variability.