This is a two-stage soft shadow visualization technique for deformable moving area light sources using adaptively sampled light field shadow maps. In the preprocessing stage, we create a slab of shadow maps for the sample points in a light template. To reduce storage cost, our adaptive sampling method stores each map in a quadtree and only stores fine details from those areas that need them. In the dynamic visualization stage, we make use of the shadow maps to efficiently construct/update the shadowed image. Once the scene is processed for the initial lighting setup, a new lighting setup can be computed even more rapidly if most of the light points do not change.