These results show comparisons of novel view synthesis from our approach vs. renderings using depths from Consistent Video Depth Estimation [Luo et al 2020].