In this paper, we consider temporal aggregation of volatility models. We introduce a semiparametric class of volatility models termed square-root stochastic autoregressive volatility (SR-SARV) and characterized by an autoregressive dynamic of the stochastic variance. Our class encompasses the usual GARCH models and various asymmetric GARCH models. Moreover, our stochastic volatility models are characterized by observable multiperiod conditional moment restrictions. The SR-SARV class is a natural extension of the weak GARCH models. Our extension has four advantages: i) we do not assume that the fourth moment is finite; ii) we allow for asymmetries (skewness, leverage effect) that are excluded by the weak GARCH models; iii) we derive conditional moment restrictions which are useful for non-linear inference; iv) our framework allows us to study temporal aggregation of IGARCH models and non-linear models such as EGARCH and Exponential SV in discrete and continuous time.