Video reconstruction has become an important tool for rig and artefact removal in cinema postproduction. In this paper we are concerned with reconstructing stereo video material. We propose a method that builds on existing exemplar-based video inpainting techniques and includes a dedicated view consistency constraint. Within a constrained texture synthesis framework, we use reconstructed motion and inter-frame disparity vectors as guides for finding appropriate example source patches from parts of the sequence that minimise spatial and stereo discrepancies. We then introduce coherent patch sewing to reconstruct the missing region by stitching the source patches together. Compared to previous methods our results show increased spatial and view consistency.