A macroscopic loading model for multi-directional, time-varying and congested pedestrian flows is proposed. Walkable space is represented by a network of streams that are each associated with an area in which they interact. To describe this interaction, a stream-based pedestrian fundamental diagram is used that relates density and walking speed in multidirectional flow. The proposed model is applied to two different case studies. The explicit modeling of anisotropy in walking speed is shown to significantly improve the ability of the model to reproduce empirically observed walking time distributions. Moreover, the obtained model parametrization is in excellent agreement with the literature. (C) 2016 Elsevier Ltd. All rights reserved.